Рубрика: AI

AI world news.

How tech professionals can survive and thrive at work in the time of AI

Male computer programmer working late in the office on a new code

Despite a spate of fresh headlines about tech layoffs, opportunities for tech professionals are robust — it's a matter of adapting to a fusion of new technologies and relentless business requirements. Overall employment of software developers, quality assurance analysts, and testers is projected to grow 25% from 2022 to 2032, "much faster than the average for all occupations," states an analysis out of the U.S. Bureau of Labor Statistics, issued in September 2023.

Also: Want to work in AI but don't know where to start? Follow these 5 steps

"Increased demand for software developers, software quality assurance analysts, and testers will stem from the continued expansion of software development for artificial intelligence, Internet of Things, robotics, and other automation applications," according to the BLS statement.

But the potential quantity of tech-related opportunities is only one side of the story. What's really compelling is the quality of tech-related work we will be seeing in the very near future. Technologies such as AI and low and no-code platforms may be pushing away manual work in favor of higher-level tasks. Even those currently working with AI will be in a position to expand their skill bases.

Changes in skill requirements are borne out by research from LinkedIn's Economic Graph Research Institute. "We're in a period of rapid and continuous change in the skills required to perform our jobs," says Dan Brodnitz, global head of content strategy at LinkedIn Learning. Insights from LinkedIn's data suggest "more than half of LinkedIn members hold jobs that stand to be either disrupted or augmented by AI, and the skills required for our jobs will change by up to 65% by 2030," he relates.

With the onset of artificial intelligence and machine learning in day-to-day work, "certain technology-related skills are facing a paradigm shift," says Harshul Asnani, president of enterprise technology at Tech Mahindra. These include anything involving "repetitive and rules-based tasks, traditionally handled by humans. Basic data entry, routine coding for standard applications, and even some aspects of rudimentary data analysis are becoming automated through AI algorithms."

Also: How to use Image Creator from Microsoft Designer (formerly Bing Image Creator)

In addition, over time, skills such as "code abstraction, code conversion, creating visual artifacts through coding, and basic code testing and QA are likely to become less relevant," predicts Asnani.

This does not necessarily mean such skills "will be eliminated, but rather transformed," he continues. Tech professionals "will need to adapt by focusing on more complex, creative, and strategic tasks that AI cannot easily replicate."

Technology-related skills coming to the forefront include "machine learning, data structures, and natural language processing," says Brodnitz. "These technical proficiencies form the backbone of AI applications and are crucial for efficient data handling, learning algorithms, and language-based AI interactions. It's important for IT professionals aiming to stay competitive in today's job market to build and strengthen these skills."

Also: Have 10 hours? IBM will train you in AI fundamentals — for free

Tech professionals "who understand how to think bigger and apply AI to make progress against strategic goals will be well-positioned to succeed," says Joe Bradley, chief scientist at LivePerson. "For example, customer service, sales, and marketing functions are increasingly supported by AIs that are customers' first lines of contact, staffing the digital front door. Just like websites are constantly optimized, customer-facing AI can be continually improved to support the goals of service, sales, and marketing leaders."

Important tools or platforms for designing, building, and managing 2020s solutions include "GitHub, Slack, Hugging Face, Reddit, and other existing open-source collaboration platforms," says Asnani. "These resources are useful in accelerating their learning by leveraging the technical knowledge and code contributions of others."

Along with designing and building AI, there is rising demand for human oversight to ensure that technologies deliver results that are trustworthy and meaningful to the business. "Fundamentally, AI and ML technologies deal with data; and to that end, people with data management and science skills will be in high demand," says Srini Kadiyala, chief technology officer at OvalEdge. "People must curate the specific data sets and sources required to fuel AI algorithms and learning models."

Also: I spent a weekend with Amazon's free AI courses, and highly recommend you do too

Data preparation "is currently one of the most essential requirements for the efficient running and output from AI modules," Kadiyala observes. "While there is a degree of autonomy, in that AI can complete data preparation tasks, to ensure accuracy and compliance, it is still essential to have a trained professional take the reins in this area."

It's all about the increasing "democratization of both learning and technological development," Asnani states. "Professionals should cultivate the ability to learn, unlearn, and relearn quickly. Their learning scope should also encompass data structures and algorithms, data analysis, mathematics, and software engineering."

Tech professionals' jobs have already gone through major changes with websites, mobile apps, and social media, says Bradley. "They're going to be changed again, reshaped by the people who know where AI will be most effective, where humans need to hold the reins."

Also: 6 AI tools to supercharge your work and everyday life

Along with skills demands, new types of job titles are emerging. The LinkedIn platform has seen "an increase in AI-related titles," Brodnitz says. "For example, over the last five years, the number of companies featuring a dedicated 'head of AI' has more than tripled, indicating the growing recognition of AI's significance within organizations."

Additional emerging roles include "AI ethics specialist, smart contracts architect, blockchain network deployer, quantum computing engineer, and VR experience designer," says Asnani. "There is also a growing need for data privacy managers in response to heightened focus on data security and privacy regulations."

As organizations lean ever more heavily on their tech talent to deliver business results and growth, expect to see more movement away from heads-down programming and technical roles. As part of this trend, low-code and no-code platforms will define technology work in the year ahead. The growth of these technologies "is set to accelerate dramatically, propelled by advancements in AI and machine learning," says Asnan.

Also: I took this free AI course for developers in one weekend and highly recommend it

This shift means "IT professionals will be relieved from the tasks of writing basic code, conducting code testing, and performing quality assurance." says Asnani. "Instead, their focus will shift towards validating the outcomes of code and ensuring they meet desired objectives and results. This transition necessitates a deeper understanding of specific domains and enhanced functional knowledge."

Going forward, success in technology work hinges on "intellectual fearlessness and curiosity," Bradley says. "Don't get locked into one way of thinking. Seek alternate approaches and perspectives. "Whatever your expertise, don't worry if it perfectly applies to your chosen field, but instead go deep, do it, and enjoy it."

Even if one is adept at AI development, there is still a lot to learn, he continues. For example, "Iyou're a machine learning expert, take a challenging course in product marketing. Don't cede absolute authority to other experts. Know enough and be curious enough to question them."

Featured

New MIT CSAIL study suggests that AI won’t steal as many jobs expected

New MIT CSAIL study suggests that AI won’t steal as many jobs expected Kyle Wiggers 11 hours

Will AI automate human jobs, and — if so — which jobs and when?

That’s the trio of questions a new research study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), out this morning, tries to answer.

There’s been many attempts to extrapolate out and project how the AI technologies of today, like large language models, might impact people’s’ livelihoods — and whole economies — in the future.

Goldman Sachs estimates that AI could automate 25% of the entire labor market in the next few years. According to McKinsey, nearly half of all work will be AI-driven by 2055. A survey from the University of Pennsylvania, NYU and Princeton finds that ChatGPT alone could impact around 80% of jobs. And a report from the outplacement firm Challenger, Gray & Christmas suggests that AI is already replacing thousands of workers.

But in their study, the MIT researchers sought to move beyond what they characterize as “task-based” comparisons and assess how feasible it is that AI will perform certain roles — and how likely businesses are to actually replace workers with AI tech.

Contrary to what one (including this reporter) might expect, the MIT researchers found that the majority of jobs previously identified as being at risk of AI displacement aren’t, in fact, “economically beneficial” to automate — at least at present.

The key takeaway, says Neil Thompson, a research scientist at MIT CSAIL and a co-author on the study, is that the coming AI disruption might happen slower — and less dramatically — than some commentators are suggesting.

“Like much of the recent research, we find significant potential for AI to automate tasks,” Thompson told TechCrunch in an email interview. “But we’re able to show that many of these tasks are not yet attractive to automate.”

Now, in an important caveat, the study only looked at jobs requiring visual analysis — that is, jobs involving tasks like inspecting products for quality at the end of a manufacturing line. The researchers didn’t investigate the potential impact of text- and image-generating models, like ChatGPT and Midjourney, on workers and the economy; they leave that to follow-up studies.

In conducting this study, the researchers surveyed workers to understand what an AI system would have to accomplish, task-wise, to fully replace their jobs. They then modeled the cost of building an AI system capable of doing all this, and also modeled whether businesses — specifically “non-farm” U.S.-based businesses — would be willing to pay both the upfront and operating expenses for such a system.

Early in the study, the researchers give the example of a baker.

A baker spends about 6% of their time checking food quality, according to the U.S. Bureau of Labor Statistics — a task that could be (and is being) automated by AI. A bakery employing five bakers making $48,000 per year could save $14,000 were it to automate food quality checks. But by the study’s estimates, a bare-bones, from-scratch AI system up to the task would cost $165,000 to deploy and $122,840 per year to maintenance… on the low end.

“We find that only 23% of the wages being paid to humans for doing vision tasks would be economically attractive to automate with AI,” Thompson said. “Humans are still the better economic choice for doing these parts of jobs.”

Now, the study does account for self-hosted, self-service AI systems sold through vendors like OpenAI that only need to be fine-tuned to particular tasks — not trained from the ground up. But according to the researchers, even with a system costing as little as $1,000, there’s lots of jobs — albeit low-wage and multitasking-dependent — that wouldn’t make economic sense for a business to automate.

“Even if we consider the impact of computer vision just within vision tasks, we find that the rate of job loss is lower than that already experienced in the economy,” the researchers write in the study. “Even with rapid decreases in cost of 20% per year, it would still take decades for computer vision tasks to become economically efficient for firms.”

The study has a number of limitations, which the researchers — to their credit — admit. For example, it doesn’t consider cases where AI can augment rather than replace human labor (e.g. analyze an athlete’s golf swing) or create new tasks and jobs (e.g. maintaining an AI system) that didn’t exist before. Moreover, it doesn’t factor in all the possible cost savings that can come from pre-trained models like GPT-4.

One wonders whether the researchers might’ve felt pressure to reach certain conclusions by the study’s backer, the MIT-IBM Watson AI Lab. The MIT-IBM Watson AI Lab was created with a $240 million, 10-year gift from IBM, a company with a vested interest in ensuring that AI’s perceived as non-threatening.

But the researchers assert this isn’t the case.

“We were motivated by the enormous success of deep learning, the leading form of AI, across many tasks and the desire to understand what this would mean for the automation of human jobs,” Thompson said. “For policymakers, our results should reinforce the importance of preparing for AI job automation … But our results also reveal that this process will take years, or even decades, to unfold and thus that there is time for policy initiatives to be put into place. For AI researchers and developers, this work points to the importance of decreasing the costs of AI deployments and of increasing the scope of how they can be deployed. These will be important for making AI economically attractive for firms to use for automation.”

3 Interesting Uses of Python’s Context Managers

3 Interesting Uses of Python's Context Managers
Image by johnstocker on Freepik

A while ago, I wrote a tutorial on writing efficient Python code. In it, I talked about using context managers and the with statement to manage resources efficiently.

I used a simple file handling example to show how files are automatically closed when the execution exits the with block—even if there is an exception.

While file handling is a good first example, it can quickly get boring. That is why I'd like to go over other interesting uses of context managers—beyond file handling—in this tutorial. We’ll focus on handling database connections, managing subprocesses, and high-precision floating point arithmetic.

What Are Context Managers in Python?

Context managers in Python allow you to write cleaner code when working with resources. They provide a concise syntax to set up and tear down resources through:

  • An enter logic that gets called when the execution enters the context and
  • An exit logic the gets called when the execution exits the context

The simplest example of this is in file handling. Here we use the open() function in the with statement to get a file handler:

with open('filename.txt', 'w') as file:      file.write('Something random')

This acquires the resource—the file object—that is used (we write to the file) within the code block. The file is closed once the execution exits the context; so there are no resource leaks.

You can write the generic version of this like so:

with some_context() as ctx:      # do something useful on the resource!    # resource cleanup is automatic

Now let’s proceed to the specific examples.

1. Handling Database Connections

When you're building Python applications, it's quite common to connect to databases and query the tables they contain. And the workflow to do this will look like so:

  • Install the database connector to work with the database (such as psycopg2 for Postgres and the mysql-connector-python for MySQL databases).
  • Parse the config file to retrieve the connection parameters.
  • Use the connect() function to establish connection to the database.

3 Interesting Uses of Python's Context Managers
Connecting to the db | Image by Author

Once you’ve connected to the database, you can create a database to query the database. Run queries and fetch the results of the query using the run and fetch cursor methods.

3 Interesting Uses of Python's Context Managers
Querying the db | Image by Author

In doing so, you create the following resources: a database connection and a database cursor. Now let’s code a simple generic example to see how we can use the connection and the cursor objects as context managers.

Parsing TOML Files in Python

Consider a sample TOML file, say db_config.toml, containing the required info to connect to the database:

# db_config.toml    [database]  host = "localhost"  port = 5432  database_name = "your_database_name"  user = "your_username"  password = "your_password"

Note: You need Python 3.11 or a later version to use tomllib.

Python has a built-in tomllib module (introduced in Python 3.11) that lets you parse TOML files. So you can open the db_config.toml file and parse its contents like so:

import tomllib    with open('db_config.toml','rb') as file:  	credentials = tomllib.load(file)['database']

Notice that we tap into the ‘database’ section of the db_config.toml file. The load() function returns a Python dictionary. You can verify this by printing out the contents of credentials:

print(credentials)
Output >>>  {'host': 'localhost', 'port': 5432, 'database_name': 'your_database_name', 'user': 'your_username', 'password': 'your_password'}

Connecting to the Database

Say you want to connect to a Postgres database. You can install the psycopg2 connector using pip:

pip install psycopg2

You can use both the connection and the cursor objects in with statements as shown:

import psycopg2    # Connect to the database  with psycopg2.connect(**credentials) as conn:  	# Inside this context, the connection is open and managed    	with conn.cursor() as cur:      	# Inside this context, the cursor is open and managed        	cur.execute('SELECT * FROM my_table')      	result = cur.fetchall()              print(result)

In this code:

  • We use the with statement to create a context for managing the database connection.
  • Inside this context, we create another context to manage the database cursor. The cursor is automatically closed when exiting this inner context.
  • Because the connection is also closed when exiting the outer context, this construct ensures that both the connection and cursor are properly managed—reducing the chance of resource leaks.

You can use a similar construct when working with SQLite and MySQL databases too.

2. Managing Python Subprocesses

Python’s subprocess module provides functionality to run external commands inside a Python script. The subprocess.Popen() constructor creates a new subprocess. Which you can use in a with statement like so:

import subprocess    # Run an external command and capture its output  with subprocess.Popen(['ls', '-l'], stdout=subprocess.PIPE, text=True) as process:  	output, _ = process.communicate()  	print(output)

Here, we run the Bash command ls -l command to long list the files in the current directory:

Output >>>    total 4  -rw-rw-r-- 1 balapriya balapriya   0 Jan  5 18:31 db_info.toml  -rw-rw-r-- 1 balapriya balapriya 267 Jan  5 18:32 main.py

The resources associated with the subprocess are freed once the execution exits the context of the with statement.

3. High-Precision Floating-Point Arithmetic

The built-in float data type in Python is not suitable for high-precision floating-point arithmetic. But you do need high precision when working with financial data, sensor readings, and the like. For such applications, you can use the decimal module instead.

The localcontext() function returns a context manager. So you can use the localcontext() function in the with statement, and set the precision for the current context using as shown:

from decimal import Decimal, localcontext    with localcontext() as cur_context:      cur_context.prec = 40      a = Decimal(2)      b = Decimal(3)      print(a/b)

Here’s the output:

Output >>>  0.6666666666666666666666666666666666666667

Here, the precision is set to 40 decimal places—but only within this with block. When the execution exits the current context, the precision is restored to the default precision (of 28 decimal places).

Wrapping Up

In this tutorial, we learned how context managers can be used for handling database connections, managing subprocesses and contexts in high-precision floating-point arithmetic.

In the next tutorial, we’ll see how we can create custom context managers in Python. Until then, happy coding!

Bala Priya C is a developer and technical writer from India. She likes working at the intersection of math, programming, data science, and content creation. Her areas of interest and expertise include DevOps, data science, and natural language processing. She enjoys reading, writing, coding, and coffee! Currently, she's working on learning and sharing her knowledge with the developer community by authoring tutorials, how-to guides, opinion pieces, and more.

More On This Topic

  • Context, Consistency, And Collaboration Are Essential For Data…
  • New Online MS in Business Analytics for Managers from University of…
  • Qualities Hiring Managers Are Looking For in Data Scientists
  • How LinkedIn Uses Machine Learning To Rank Your Feed
  • Top Programming Languages and Their Uses
  • KDnuggets News, November 16: How LinkedIn Uses Machine Learning •…

How to add a new credential to your LinkedIn profile, and why you should

linkedinfingergettyimages-1708753972

Professional certificates and credentials can provide a nice little boost to your career prospects. They tell prospective employers that you took the time to learn a skill or focus on an area of study, to the point where you were competent enough to pass an evaluation.

Don't confuse certificates and credentials with accredited degrees. Degrees are a far longer (and usually far more expensive) course of study. In many cases, degrees show employers that you have foundational learning and discipline in your area of study, whereas credentials show employers you have studied a specific skill or topic.

Also: Open to work? Three new LinkedIn features make job searching easier than ever

Watch out for training institutions that hawk so-called "micro-degrees." They claim they're giving you a degree for a few months' worth of work. A micro-degree is not a full degree. It's just another name for a credential.

All that said, let's say you just completed a credential. How do you show it off on LinkedIn? That's what this article is going to show you.

I've been hearing from many of you that you took the IBM AI Fundamentals certificate program I wrote about last month. Good for you! It's a pretty good use of time, with some solid training and not all that much brand-specific hype.

Like you, I decided to take the full certificate program myself. I did it over the holiday break. Using my completion process, I'll show you how you can add a certificate to your LinkedIn profile.

How to add a credential to your LinkedIn profile

What you'll need: The only thing you'll need for this is a LinkedIn account.

Also: I took this free AI course for developers in one weekend and highly recommend it

Credly provides several social sharing options, as well as a way for you to print out or share a more formal verified certificate so that a prospective employer has proof you earned the credential.

Now, you're ready to add that credential to your LinkedIn profile.

On your main profile page, click the Add Profile Section button, located right under your main header.

LinkedIn will present a list of possible profile sections you can add. If you don't see licenses & certifications, hit the drop-down arrow in the Recommended section to reveal more options.

Click "Add licenses & certifications."

Fill it out appropriately. If this isn't your first credential, scroll down to your existing credentials and click the + icon.

And there you go. Your credential is now on LinkedIn for all to see.

Also: Your next employer is more interested in your skills than your degrees

What about you?

Have you taken the IBM AI course? Have you taken any other AI courses? What other credential programs have you participated in? What did you think of the learning experience? Let us know in the comments below.

You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter on Substack, and follow me on Twitter at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, and on YouTube at YouTube.com/DavidGewirtzTV.

Featured

Turn headwinds into opportunity in 2024

Turn headwinds into opportunity in 2024 Danny Rimer 9 hours Danny Rimer Contributor Share on X Danny Rimer is an investor and partner at Index Ventures.

The founders I work with know I think about John Coltrane a lot. Lately, I’ve been thinking about how he transformed jazz with a harmonic progression known as “Coltrane changes.”

Popularized on his 1960 album “Giant Steps,” Coltrane changes are characterized by rapid and frequent modulations between key centers. Breaking the mold of traditional jazz improvisation, the complex progressions challenged musicians to explore new scales and patterns to navigate the changes. They influenced the evolution of jazz as we know it today.

What does any of this have to do with starting a business? In a year like 2023, a lot.

In the business world, 2023 was a year when companies had to go back to basics and adapt their strategies to a volatile macroeconomic environment.

For founders, that meant rethinking the way they were building and growing. It meant seeing cash on the balance sheet as a static object — the thing required to stay alive. It meant making tough personnel choices, thinking hard about who was indispensable and choosing expertise over loyalty. In an uneasy market still awaiting the full impact of AI, it meant doing everything necessary to ensure their product’s place as a must-have and not a nice-to-have.

For investors, too, it was a year of extremes. On one hand, you had the AI frenzy, with everyone rushing to create the next great AI company. On the other hand, many would-be entrepreneurs remained on the sidelines, either because they had been burned by crypto or thought fundraising would be too difficult.

I’ve tried to be a voice of reason in my conversations with founders. Adaptability is essential, and startups are a marathon, not a sprint. We can look at past downturns and say they give rise to some of the best companies and leaders. In the same way, “Giant Steps” challenged musicians to innovate to keep up with Coltrane’s rapid changes.

This year, 2024, is a time for entrepreneurs to get creative and build the resilience, skills, and discipline that will carry them through the next 20 years.

Get ready for the next wave of generational startups

We’ve seen it throughout history: In economic downturns, when it’s hard to raise money, the best entrepreneurs step up.

Entrepreneurship is all about taking risks. . . . It means innovating without fear of failure, stepping into the unknown, and pursuing ambitious ideas.

If you think of the most innovative and successful startups of the past 20 years, many of today’s household names — Stripe, Uber, Airbnb, and Square — emerged after the 2008 financial crisis. Led by visionary founders, these companies seized on ideas that they believed could disrupt traditional markets and industries, operating with a focus, discipline, and entrepreneurial spirit that becomes a superpower in times of scarcity.

Dropbox had nine employees in 2008 when the company raised its Series A. Not only did Drew Houston have a clear vision of how cloud storage would transform how people store files and collaborate, but he also operated with a scarcity mindset that helped the company be more creative and efficient in allocating resources. By the time we led Dropbox’s Series B in 2011, the company had more than 45 million users, despite adding only a handful of employees.

In 2024, I believe we’ll see a similar cohort of generational founders emerge. The most successful ones will be those with the strongest core beliefs and conviction, who operate with discipline, focus, and dedication to the task at hand, and who can tell a compelling story that convinces talented people to join them on their journey.

AI will be at the forefront of that wave — led by visionary entrepreneurs

AI will continue to dominate headlines in 2024. However, I’m most interested in seeing how AI technology gets productized and commercialized and how entrepreneurs think about applying it to everyday business applications.

Since ChatGPT shocked the world a year ago, there’s been such a firestorm of enthusiasm around AI that it can be hard to separate the practical potential from the hype. But already, we’re seeing the dust start to settle, and new companies are popping up with a real entrepreneurial focus on how AI can be harnessed to create relevant products and services.

That trend will only accelerate in 2024, as every company develops its AI strategy and begins to incorporate AI into its workflows. This paradigm shift will open the door for a new wave of market disruption, bringing AI out of the realm of hype and establishing it as the foundation for the next wave of genuinely innovative startups.

I’m particularly interested in seeing how the next wave of ambitious entrepreneurs attack this opportunity. Remember that in the early days of AI, innovation was led primarily by researchers at academic institutions. These groups have done an incredible job of bringing us to where we are today and will continue to play a pivotal role as technology develops at a rapid pace. But there’s a difference between innovating in a lab to solve a complex technical problem and creating a product that delivers value to a well-defined market.

When we invested in Cohere two years ago, we did so because we loved its founders’ approach to productization. While Aidan, Ivan, and Nick were bona fide researchers and had learned under academic giants like Geoffrey Hinton (“the godfather of AI”), they also had a unique vision of how to productize large language models to help enterprise companies build practical, everyday business applications.

We felt the same when we led biotech startup Cradle‘s seed and Series A rounds. Not only do Stef and his co-founders have a rare blend of deep machine learning expertise and protein engineering experience from top tech and biotech firms, but they’ve also uncovered a strong appetite for their product among R&D teams, with massive upside given the market scale.

We’re still in the early innings of AI development. Much like Yahoo laying a path for Google, or MySpace paving the way for Facebook, AI will need time to reach its final form. Currently, visionary founders are studying and learning from developments in AI, getting ready to create the next wave of generational companies.

Dormant sectors are in for an AI awakening

One of my favorite and most surprising takeaways of 2023 was getting to see specific sectors in a new light thanks to the promise of AI. Moving forward, that will only continue to accelerate.

Advertising is a perfect example of this. It’s been a while since we saw any breakthroughs in ad technology. Still, with targeting and personalization getting more accessible and more sophisticated thanks to AI, plus the still relatively untapped potential of predictive analytics and programmatic advertising, I think we’re about to see big changes in that industry.

Dating is another sector that could use a new wave of disruption. As we all know, dating is a deeply personal human experience. Online dating has enabled connection, but it has also introduced challenges. Critics may argue that adding AI will dehumanize dating apps. Still, I see the opposite: Whether it’s better matching algorithms, more personalized recommendations, a more secure user experience, or even features that tap into augmented or virtual reality, these applications could allow people to focus more on human connection. There’s an opportunity for whoever can strike the right balance to take the lead in this sector.

And then there are all the other sectors I’ve long been excited about, which I think are primed for innovation — the creator class, the gaming industry, personal productivity apps. I’m fascinated to see how AI takes these sectors to new levels in 2024 and to witness new leaders emerge.

Regulating AI will be a global responsibility

I’m the furthest thing from a nationalist, and I find it strange when I see nation-first rhetoric seeping into startup culture.

AI is a massively transformational technology with real risks that are already starting to emerge. Of course, we need to be thoughtful about how it’s deployed, but talking about these complex issues in nationalistic terms is a distraction from the core objective — ensuring that these technologies are applied ethically and safely. Getting this right will take global collaboration.

Remember that most AI technologies transcend national borders; the companies that develop and deploy them operate globally, which means their impact extends across jurisdictions. From one country to the next, differences in national approach will lead to fragmentation and inconsistencies, exposing vulnerabilities, sapping innovation, and creating a patchwork of regulations that are less than the sum of their parts.

While geopolitical differences may make regulation more complex and challenging globally, a global approach is the only way to put adequate guardrails around AI’s safe and ethical use and ensure a landscape where AI innovation can thrive. The conversation must shift from regulating the core technology based on a hypothetical threat of AI apocalypse to addressing the actual use cases and threats emerging today.

So, how should founders think about turning headwinds into opportunities? The best entrepreneurs find a way to tune out the noise and execute their vision as only they can.

The days of “low-risk, high-reward” are gone

Thanks to historically low interest rates, a generation of entrepreneurs have been tricked into believing big rewards are possible without risk — that you can float to the top of the mountain on a magic carpet made of money. I’m sorry, but that was a mirage.

Entrepreneurship is all about taking risks. And I don’t mean incremental risk — real, transformative risk. That means innovating without fear of failure, stepping into the unknown, and pursuing ambitious ideas. It means making bets with a growth mindset, turning failure into resilience, and being bold enough to continue trying things that aren’t guaranteed to work.

Slack co-founder Stewart Butterfield knows this better than almost anyone else. Not once, but twice in his career, Butterfield has had the conviction to build a massively multiplayer online role-playing game — and both times, when he realized his experiments were failing, he had the courage to pivot. In the first case, what began as a sharable in-game photo inventory later became Flickr, which Butterfield sold to Yahoo barely 12 months after its official launch.

A similar story unfolded a few years later when Butterfield shut down his second game, Glitch, after realizing it wouldn’t make any money. His company, which had raised $15 million to develop Glitch, pivoted to focus on an internal communication tool they were building. The rest of the story needs no telling: Within two years of its public release, Slack had raised $340 million, attracted more than 2 million daily active users, and been named Inc.’s 2015 Company of the Year. Five years later, Salesforce acquired Slack for $27.7 billion.

Founders who choose low-risk paths are disadvantaged compared to competitors who are willing to take risks and innovate more aggressively. As an investor, I’ll always back the founder who believes in their vision and who is willing to make the big bet that others might shy away from because that’s where you find the best returns.

As for failure? When you dream big, it’s inevitable. The important thing is to learn from your failures. Remember Samuel Beckett’s words: “Try again. Fail again. Fail better.”

Discipline is more important than big valuations

In my experience — and I tell this to founders all the time — a company’s success is often inversely proportional to the amount of money raised in their first round.

When I look at our portfolio companies, some of the biggest success stories started with humble beginnings. Datadog, with a current market cap of $38 billion, raised $6.2 million in its Series A round. Figma began with $3.9 million in seed funding. Discord started with $1.1 million. Roblox‘s Series A was all of $560,000.

These companies and their founders are great examples of how an early scarcity mindset can instill discipline — one of the most important qualities any entrepreneur can have — and strip away distractions and optionality to do anything but what’s vital to business success.

When we met Adyen‘s founders, Pieter and Arnout, in 2011, we were immediately sold on their vision of creating a global payments solution. Ambitious? Sure, especially for a small Dutch company in a highly regulated industry. But the company was already profitable, with customers signed up across four continents. They were so disciplined they didn’t need our money, and it was on us to convince them to let us lead their Series A.

As investment picks up in 2024, I’m sure we’ll see some jaw-dropping valuations. Refrain from overthinking these big valuations automatically translate into success. Just as we’ve seen many successful companies start with humble beginnings, I can think of plenty of companies that raised huge first rounds and failed due to a lack of discipline, internal challenges, or just plain getting outplayed by the competition.

Don’t sacrifice growth for profitability at all costs

If you talk to the folks on Wall Street, they’ll tell you that profitability is all that matters. But you can’t run your business based on what Wall Street wants. That’s the business equivalent of letting the tail wag the dog.

Of course, profitability is essential, but you shouldn’t choose short-term efficiency at the expense of long-term ambition. This goes back to having a vision and a willingness to take risks. The most successful companies are the ones that can grow profitably with increased margins and efficiency. The first part of that equation is figuring out how to drive growth.

In 2023, no one would have criticized Figma for doing another small developer conference. But with all eyes on them in the wake of the since-abandoned Adobe acquisition and no one else advertising or investing in big developer conferences, they saw their opportunity. They took a risk and held their biggest conference ever. And guess what? It was a massive success, with more than 8,500 attendees. It completely changed how Figma is perceived in the market, giving them a proven lever they can pull in future years to drive even more growth.

It always comes back to basics

As humans, we’re addicted to newness, but newer isn’t always better. Bigger isn’t always better. And even if something is different or exciting, there’s still a market for it.

The world is changing faster than ever. The innovation in 2024 will be unlike anything we’ve seen in history. I’m excited about it, but I’m also mindful of not getting carried away by the hype. Whether you’re a founder or investor, we need to remember that the core ingredients of a successful business have stayed the same:

  • Visionary leadership.
  • A clear value proposition.
  • A well-defined market.
  • A product or service that provides real value.

These principles gave us the confidence to invest in Figma in 2013. When I met Dylan, he was a 19-year-old intern at LinkedIn. There was no reason anyone could find on paper to invest in him and Evan. But we believed in their vision, and more importantly, we believed in their conviction to build the most important product design company in the world.

At Index, we’ve always been transparent about our focus on investing in people. Building a business is a craft; the entrepreneur is the ultimate craftsperson. As investors, we do what we can to empower and support them, but the entrepreneur is the central figure and the only person who knows what’s best for their business.

The companies that are most successful in 2024 will be the ones that reflect the true spirit of entrepreneurship, which is all about having big ambitions, a compelling vision, and total dedication to the cause. I’m excited to see who emerges and what their vision looks like and to do our part by supporting them on their journey.

Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language Model

Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language Model
Photo by Google DeepMind

2023 was the year of Large Language Models and Open Source. Many startups and companies open-sourced their models and weights to combat proprietary LLMs such as ChatGPT and Claude. Some of the important companies and models (open source) for 2023 were:

  • Meta (LLama, LLamav2)
  • TII (Falcon 7B, 40B, 180B)
  • Mistral (Mistral 7B, Mixtral8x7B)

However, a 7B model which is relatively easy and cheaper to deploy is not up to par with bigger models such as 70B. The strongest open-source contender was Mistral 7B which would outperform many bigger models.

Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language Model
Comparison of Mistral-7B from Mistral.ai

These small models however still do not respond well to natural prompts and require good prompt engineering.

Introduction

Zephyr 7B is a model created by the HuggingFace H4 (Helpful, Honest, Harmless, Huggy) team whose main goal was to create a smaller language model that is aligned with user intent and outperforms even bigger models.

Zephyr is an aligned version of Mistral-7B mainly created with the power of Distillation, and is comparable to 70B models in academic and conversational benchmarks.

Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language ModelPerformance comparison of Zephyr-7B | Source: Zephyr paper

Key Features

The reason behind the outstanding performance of Zephyr is these 4 key techniques that the H4 Team has used.

  1. Self-Instruct data creation & DSFT (Distilled Supervised Fine-Tuning)
  2. Feedback collection
  3. DDPO (Distilled Direct Preference Optimization) of the DSFT model

Self-Instruct Data Creation & DSFT

Traditionally Supervised Fine-Tuning (SFT) is performed on a Large Language Model via a high-quality instruction completion pair. Construction of this data is costly and requires human supervision (Chung et al., 2022; Sanh et al., 2021).

One of the interesting approaches here is to use a Teacher model (already trained LLM) to generate the instructions and responses. This distillation technique was first used on Alpaca (Taori et al., 2023) which proved that a small model can outperform larger models with Distilled Supervised Fine-Tuning.

Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language Model
Self-Instruct pipeline | Source: Self-Instruct paper

The H4 Team used Zephyr for constructing high-quality supervised (instruction, completion) datasets that were used for doing DSFT. (Training a model on instructions/completions generated is a form of distillation known as DSFT: Distilled Supervised Fine-Tuning).

Feedback Collection

Large Language Models are aligned typically with the help of Reinforcement learning from human feedback (RLHF). Zephyr instead uses Feedback from a better teacher model (such as GPT-4) to align the interests of the model, following the approach of Ultra Feedback.

Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language Model
UltraFeedback construction process | Source: UltraFeedback paper

The way it works is that each prompt supervised prompt from SFT is passed to 4 models (Claude, LLama, Falcon, etc.) and each of the 4 responses against the single prompt is scored with the help of GPT-4. Now we have a dataset of an Input (x), highest scoring completion (yw), and a random prompt denoted as low scoring completion (yl), i.e we have a triplet of (x, yw, yl).

Preference Optimization

The goal of this last step is to maximize the preference of the model from yw(highest-scoring completion) over yl (low-scoring completion). This is done using DPO (Direct Preference Optimization). Using DPO is simpler than using plain RLHF and intuitively it performs better than RLHF. The approach in this case is known as dDPO because it uses a distilled dataset generated with the help of a teacher model.

Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language ModelDPO vs RLHF | Source: Zephyr paper

The overall algorithm looks somewhat like this:

Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language Model

And can be translated into the following steps:

  1. Compute the probability for (x, yw) and (x, yl) from the dSFT model (forward-only).
  2. Compute the probability for (x, yw) and (x, yl) from the dDPO model.
  3. Compute Eq 1 and backpropagate to update. Repeat

Training Details

The base model that Zephyr used is Mistral-7B which was the state-of-the-art open source at the time of release. They used the TRL library for fine-tuning and alignment. Deep-Speed Zero 3 and Flash-Attention 2 were used to optimize and speed up the training and to fully utilize the GPU. The models were trained using AdamW optimizer and no weight decay was used. All experiments were run on 16 A100s using bfloat16 precision and typically took 2–4 hours to complete. You can refer to the original paper for in-depth details on the Training Procedure of Zephyr.

Results

Zephyr team combines the best techniques to train the Large Language Models and it matched the performance of 40B models with just 7B parameters and matched 70B for chat models.

Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language ModelComparison of Zephyr vs other LLMs | Source: Zephyr paper
Exploring the Zephyr 7B: A Comprehensive Guide to the Latest Large Language ModelComparison of Zephyr vs other LLMs | Source: Zephyr paper Usage

Zephyr models are publically available on Hugging Face and can be used similarly to any other Language Model.

import torch  from transformers import pipeline    pipe = pipeline("text-generation",                  model="HuggingFaceH4/zephyr-7b-alpha",  # can also use the beta model                  torch_dtype=torch.bfloat16,                  device_map="auto")    # We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating  messages = [     {         "role": "system",         "content": "You are a friendly chatbot who always responds in the style of a pirate",     },     {"role": "user", "content": "How many helicopters can a human eat in one sitting?"},  ]  prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)  outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)  print(outputs[0]["generated_text"])

Output:

<|system|>  You are a friendly chatbot who always responds in the style of a pirate.  <|user|>  How many helicopters can a human eat in one sitting?  <|assistant|>  Ah, me hearty matey! But yer question be a puzzler! A human cannot eat a helicopter in one sitting, as helicopters are not edible. They be made of metal, plastic, and other materials, not food!

Conclusion

Zephyr-7B is a small model that showed the power of distillation from a LLM to a smaller model. The resulting model ZEPHYR-7B, based on MISTRAL-7B, sets a new state-of-the-art for 7B parameter chat models and even outperforms LLAMA2-CHAT-70B on MT-Bench.

References

  1. Zephyr: Direct Distillation of LM Alignment (https://arxiv.org/abs/2310.16944)
  2. HuggingFace Zephyr blog (https://huggingface.co/blog/Isamu136/understanding-zephyr)
  3. Self Instruct: https://arxiv.org/abs/2212.10560
  4. UltraFeedback: https://arxiv.org/abs/2310.01377

Ahmad Anis is a passionate machine learning engineer and researcher currently working at redbuffer.ai. Beyond his day job, Ahmad actively engages with the machine learning community. He serves as a regional lead for Cohere for AI, a nonprofit dedicated to open science, and is an AWS community builder. Ahmad is an active contributor at Stackoverflow, where he has 2300+ points. He has contributed to many famous open-source projects, including Shap-E by OpenAI.

More On This Topic

  • Exploring the Latest Trends in AI/DL: From Metaverse to Quantum Computing
  • A Comprehensive List of Resources to Master Large Language Models
  • The Ultimate Open-Source Large Language Model Ecosystem
  • Introduction to NExT-GPT: Any-to-Any Multimodal Large Language Model
  • A Comprehensive Guide to Convolutional Neural Networks
  • A Comprehensive Guide to Pinecone Vector Databases

How Pure Storage is Optimising Data Storage for Running AI Workloads

The most contentious roadblock to running AI workload is not just the efficiency and high demand for GPUs but also largely depends on optimising data storage. In an exclusive interview with AIM at its India R&D centre in Bangalore, John Colgrove, co-founder and CTO of Pure Storage, shared his perspectives on how it is helping in solving these concerns, alongside ensuring fast, accessible, and reliable data storage, leading to better resource utilisation and reducing the strain on GPUs.

“We ensure compute efficiency through scalable storage solutions that ensure data is accessible, manageable, and processed seamlessly by GPUs—through our hardware portfolio, which is GPU Direct Storage (GDS) ready and has AI-Ready Infrastructure (AIRI),” said Colgrove.

Pure Storage began its journey in California’s Bay Area in 2009, rapidly expanding to global R&D centres in Prague and Bengaluru. It expanded to India right after Covid, in June of 2022, given the opportunities and talent.

“We wanted to open a site where we could hire top-notch engineers and give them challenging and difficult work to do,” said Colgrove about the India operations—their biggest outside of the US.

Competition

While it competes with major providers like Google Cloud Storage, Amazon Elastic Block Store (Amazon EBS), and DigitalOcean in the broader storage-infrastructure market. The competition is more intense in the all-flash storage market, where Pure Storage faces giants like IBM, Hewlett Packard Enterprise (HPE), NetApp, and Dell EMC.

However, Colgrove is confident in their product, describing its 75 terabytes direct flash module as a high-performance per terabyte solution. “We have by miles the best product; the direct flash gives us a huge advantage over anyone selling hard drives,” he stated.

He also focused on the importance of power efficiency in data centres—explaining its objective to create storage solutions that are both dense and power-efficient. “We want the storage footprint to be as dense and as power-efficient as possible,” he said, acknowledging the crucial role of power management in modern data storage.

Moreover, it plans to ship a 150-terabyte version of the direct flash module later in the year. This capacity significantly surpasses traditional hard drives, with Colgrove saying, “Seagate announced like we’re going to ship 30 this year. They’re going to ship 30 This year, we’re going to ship 150 This year.”

He emphasised their direct flash drives’ ecological and economic benefits, noting their power usage and electronic waste efficiency. “There’s less e-waste. It uses less power,” Colgrove explained, highlighting the environmental advantages of its technology. He also pointed out that these drives are more cost-effective in terms of total ownership cost than traditional disks.

Colgrove also talked about the ‘Evergreen’ model employed by Pure Storage, a key differentiator in its product strategy. This approach allows the arrays to be upgraded while they’re online without data migration, making them seem new even after ten years of operation.

“The magic isn’t that they’re ten years old and still functioning; the magic is they’re ten years old, but they look like a brand new array ready for another ten years,” he said. This model eliminates the need for costly and risky data migrations common with traditional data storage solutions.

With a market share of about 6%, the company reported a revenue of $762.8 million for the third quarter of 2024. The previous fiscal year saw its revenue at $2.8 billion, a 26% increase year-over-year, with a free cash flow of $609.1 million for fiscal year 2023.

Collaborations & Partnerships

Central to Pure Storage’s strategy is its collaboration with NVIDIA in developing AI-Ready Infrastructure (AIRI) solutions. “NVIDIA is an interesting one, and we made the first AI ready infrastructure back in 2017,” Colgrove said.

The FlashBlade hardware portfolio is set to fully support GPU Direct Storage (GDS) with upcoming software updates, enhancing this collaboration.

“The AIRI//S solution, co-developed with NVIDIA and built on the NVIDIA DGX system, is designed to provide significant performance advancements that are non-disruptive, allowing for continuous innovation and adaptability to the evolving needs of AI,” added Colgrove.

The AIRI//S architecture is comprehensive and scalable, integrating FlashBlade//S, NVIDIA DGX systems, and NVIDIA networking. This integration creates an optimised setup for multi-dimensional GPU performance, ensuring GPUs are continuously engaged in AI workloads of any scale. This setup simplifies the development of modern AI environments for enterprises.

Additionally, the company plays a crucial role at various stages of AI development, starting from the initial data collection and storage phase—to the crucial phase of Data Access and Management, where its AI Storage Solutions automate the access to diverse data sources and storage resources, reducing the model training time from months to days, streamlining the entire process of data curation, training, and inference.

Pure Storage has established partnerships with most major hyperscalers. “We are working with all of the hyperscalers very closely. I think we work with almost everybody other than Google Cloud at this point,” Ajeya Motaganahalli, MD of Pure Storage R&D in India, told AIM, citing Portworx, a Kubernetes storage overlay.

He said this product functions as a container in the cloud and is compatible with various public cloud platforms, including GCP, Microsoft Azure, and AWS. It connects to underlying storage, whether on-premises or in the cloud, creating a unified namespace.

What’s next?

Going ahead, the CTO said the company’s ambition is to be a leader in the inference market while continuing to support the AI training market. “We want to be leading that, obviously, and we want to continue to support the training market as well,” he stated, saying that it aims to address both the training and application phases of AI development.

Discussing the limitations of traditional hard drives in AI, Colgrove pointed out their inadequacy for AI applications due to their slow access capacity. “Hard drives don’t have the access capacity. They’re so slow on I/O ops there… AI can’t run on top of hard drives,” he explained, highlighting the need for faster, more efficient storage solutions.

Colgrove further elaborated on the role of data in generative AI, using retail and automotive industries as examples. In retail, AI can analyse small amounts of data from video cameras in stores to understand shopper behaviour and then apply these learnings to a larger dataset.

Similarly, AI can learn from a subset of telemetry data from test vehicles in the automotive sector and then apply these insights to millions of cars. “We’re going to learn stuff by studying a subset of my data very carefully, and that’s what the training but I’m then going to apply it to millions of things,” he said, emphasising the scale at which AI operates and the consequent data requirements.

The post How Pure Storage is Optimising Data Storage for Running AI Workloads appeared first on Analytics India Magazine.

UP Police Enlists AI to Bolster Security in Ayodhya

The security measures for visitors attending the ‘pran prathishtha’ ceremony at Sri Ram Mandir in Ayodhya on January 22 have reached unprecedented levels. Undoubtedly, UP Police is relying heavily on AI to ensure smooth operations.

The police have deployed over 10,000 CCTV cameras, many of which are AI-powered. “To ensure better security arrangements at the programme venue in Ayodhya, technology is being used on a large scale. In some of these CCTV cameras, we are using AI-based technology so that we can maintain a strict vigil on the commuters,” the director general for law and order, Prashant Kumar told PTI.

These cameras have been installed in hotspots such as Kanak Bhawan, Hanuman Garhi, Shri Nageshwar Nath Mandir, Ram Ki Paidi and Ram Janmabhoomi.

In addition to the millions of devotees thronging Ayodhya, approximately 506 prominent individuals, including politicians, industrialists, film stars, sportspersons, diplomats, judges, and esteemed priests, are all attending the event.

Further elevating its significance on the global stage, the consecration ceremony will host 92 specially invited dignitaries representing 50 countries as state guests.

As a security measure, reports suggest over 15,000 police personnel, along with paramilitary forces, ATS commandos and sniper teams, have been deployed across the city.

Facial recognition tech

UP Police will use the CCTV cameras along with an AI-powered audio-video analytics platform to monitor the event for threats and suspicious activities. The platform, called Jarvis, is developed by a Gurgaon-based startup called Staqu Technologies.

JARVIS leverages AI and computer vision to analyse video footage and extract valuable insights like detecting and tracking objects, recognising faces, identifying anomalies, and performing various other video analytics tasks. It provides short and crisp real-time alerts based on the analysed video data. In Ayodhya, it will scan the cities and premises for threats and relay real-time alerts to the authorities.

The platform has been fed with a comprehensive digital database of 8,00,000 criminals in UP. The CCTV cameras will utilise advanced high-resolution facial recognition capabilities, enabling the identification and monitoring of suspects across various locations with an impressive accuracy rate of up to 99.7%, according to the startup.

The AI system will recognise and promptly alert the authorities in the event that someone from the database or any other high-risk individuals is detected in the camera feed. Additionally, these cameras are equipped with reverse facial recognition functionality, which means it can identify a person by scanning a photo of that particular person.

AI-powered number plate recognition

The cameras will also monitor the vehicles entering and exiting the city during the event. UP Police will use an advanced AI-powered Automatic Number Plate Recognition (ANPR) system for traffic management, law enforcement, and public safety.

In Ayodhya, ANPR will be specifically used by UP Police to assist in the identification of vehicles involved in criminal activities or traffic violations. Authorities can integrate ANPR with databases of wanted vehicles to identify and apprehend them automatically.

The AI system will have real-time access to the government’s vehicle registration database, which includes information from stolen vehicle databases. While only authorised vehicles are permitted to enter Ayodhya, the system can immediately detect unauthorised vehicles. Moreover, the system can also identify vehicles sporting fake number plates.

Attribute-based searches

The UP Police will harness Staqu’s JARVIS platform to enable surveillance cameras to conduct attribute-based searches. The AI-powered system can identify individuals within a crowd based on distinctive attributes like clothing, colour, accessories, or the presence of accompanying children.

Real-time monitoring using attribute-based AI searches can help manage large crowds and ensure public safety. It will help UP Police not only locate criminals but also find lost people or children in a crowd.

AI-powered anti-mine drones

Authorities are also using AI-powered drones to keep an eye on the movement of people visiting the holy place. Equipped with sensors and detection technology, these AI-driven drones are designed to scan the ground for concealed landmines or explosive devices.

“Ayodhya is now under the watchful eye of drones equipped with AI alongside the utilisation of anti-mine drones, as part of the concerted efforts to enhance security in the temple town,” a senior police official said.

The post UP Police Enlists AI to Bolster Security in Ayodhya appeared first on Analytics India Magazine.

TikTok Releases Depth Anything, Foundational Model for MDE

TikTok has unveiled Depth Anything, a groundbreaking development in the realm of Monocular Depth Estimation (MDE). This innovation harnesses the potential of a colossal dataset, comprising 1.5 million labeled images and an astonishing 62 million-plus unlabeled images.

By jointly training on such a massive scale, Depth Anything emerges as a foundational model for MDE with an array of advanced features.

Click here to check out the demo.

Key features of Depth Anything include

  • Zero-shot relative depth estimation, surpassing MiDaS v3.1 (BEiTL-512)
  • Zero-shot metric depth estimation, outperforming ZoeDepth
  • Optimal in-domain fine-tuning and evaluation on NYUv2 and KITTI datasets

Unlike previous approaches, the focus of Depth Anything is not on introducing novel technical modules. Instead, the emphasis lies on constructing a straightforward yet potent foundational model capable of handling diverse images in any scenario.

To achieve this, the dataset is scaled up significantly through the implementation of a data engine designed for collecting and automatically annotating a vast pool of unlabeled data, totaling approximately 62 million images. This extensive dataset expansion proves instrumental in reducing generalisation errors.

Two effective strategies are explored in the process:

  • Challenging Optimisation Target: A more demanding optimization target is created using data augmentation tools. This compels the model to actively seek additional visual knowledge, thereby acquiring robust representations.
  • Auxiliary Supervision: An auxiliary supervision is developed to ensure the model inherits rich semantic priors from pre-trained encoders. This enhances the model’s ability to interpret and understand images.

Extensive evaluation of Depth Anything’s zero-shot capabilities involves six public datasets and randomly captured photos, showcasing its impressive generalisation ability.

Furthermore, through fine-tuning with metric depth information from NYUv2 and KITTI, Depth Anything establishes new State-of-the-Art (SOTA) benchmarks. The improved depth model also yields superior results in depth-conditioned ControlNet.

Read: This New AI tool Could Mark the Beginning of the End for TikTok and Instagram Influencers

The post TikTok Releases Depth Anything, Foundational Model for MDE appeared first on Analytics India Magazine.

Apollo Launches India’s First AI-Precision Oncology Hub in Bengaluru

Apollo Cancer Centre has unveiled India’s first AI-Precision Oncology Centre (POC) in Bengaluru, India, aiming to improve oncology care. The center employs AI to provide timely and personalised results for oncologists, patients, and caregivers. It offers comprehensive services such as accurate diagnosis, real-time insights, cancer risk assessment, treatment protocols, and ongoing care.

The patient-centric approach includes the identification of eligible individuals for targeted therapy and immunotherapy. The team is also using conversational AI to educate patients and families on diagnosis, treatment FAQs, and connections to support groups.

“As we inaugurate India’s first AI-driven Precision Oncology Centre at Apollo Cancer Centres, Bengaluru, I am honoured to witness a groundbreaking convergence of innovation and healthcare. This pioneering initiative not only represents a monumental leap in medical technology but also symbolises hope and progress for patients and caregivers,” said Dinesh Gundu Rao, State Minister for Health & Family Welfare, Karnataka.

The center supports new patient identification through auto-alerts and aids in care pathway compliance, clinical management, and patient benefit programs.

Previously, Apollo collaborated with Microsoft to develop an AI-powered cardiac prognosis model tailored to South Asian data as traditional cardiac risk models, based on European or US data, fall short of predicting risks in the Indian population.

Even in 2019, the two companies introduced the National Clinical Coordination Committee using AI-powered risk API to combat cardiovascular diseases. Apollo has further expanded its technological initiatives, partnering with a startup for virtual reality solutions and introducing smart automation for patient monitoring.

The post Apollo Launches India’s First AI-Precision Oncology Hub in Bengaluru appeared first on Analytics India Magazine.