Рубрика: AI

In recent years, CES has gained a reputation for being a show of splashy introductions of products that will never come to market — even though the roots of CES are based in retail stores previewing the year's upcoming products to decide what to stock on their shelves. Thankfully, CES 2024 had no shortage of real product launches along with strong pitches for recently released tech gadgets.

Also: CES 2024: What's Next in Tech

ZDNET has posted its editor's pick for best of show. I thought I'd also share my personal list of products that I would pay good money for as soon as they are available to buy. These five products are at the top of my queue.

1. Roborock Flexi Pro

I've been very impressed with the quality of Roborock's robot vacuums for the past couple years, because even though they weren't first to market they have out-innovated the industry to create 2-in-1 robots that vacuum, mop, and work more reliably than their competitors. I was even more impressed after talking to Roborock's president Quan Gang at CES, where he told me that the first two products that Roborock ever made they scrapped because they weren't happy with the quality and the third product had to succeed because they were out of money. It was an instant hit and now they are one of the top-selling robovac makers globally.

Also: EcoFlow Delta Pro Ultra and Smart Home Panel 2 just launched at CES 2024 and ZDNET tested it

But the product I loved from CES wasn't a robovac. It's a smart, upright, wet/dry vacuum called the Roborock Flexi Pro (because sometimes a robovac can't get everywhere or you just need a quick cleanup). The Flexi Pro has a beautiful, minimalist design, works as both a vacuum and a mop, is self-cleaning, and has 17,000Pa of suction power (twice the power of most robovacs). The price and release date haven't been announced yet, and I can't wait to learn more.

2. Ecobee Smart Doorbell Camera

When Ecobee launched its Smart Doorbell Camera in Q4 2023, most of the industry shrugged and thought it was an odd late entrance into a market that has a ton of entrenched competitors from the likes of Ring, Google, Logitech, and Eufy and low-cost stalwarts like Blink, Wyze, and Arlo. But after getting a demo of the Ecobee Smart Doorbell Camera and talking to Ecobee CEO Greg Fyke at CES, I came away very impressed with the fact that the Ecobee team got into smart doorbells because there were specific issues with the existing products on the market and they were confident they could improve on them and deliver a better product.

Also: The best smart home tech at CES 2024

The biggest improvement is the field of view. Most doorbell cameras are great for seeing someone coming to the door, but they can't see a package that's set right in front of the door. The Ecobee doorbell has a 175-degree field of view so that it can see something that's set flush against the door. I tried this in my CES demo and it works great. I have this problem today and so I have a second camera on my porch to see packages. And since I already have an Ecobee thermostat and have been very happy with the quality, features, and usability, another nice feature of the Ecobee doorbell is that it can use the screen on your Ecobee thermostat to view the doorbell camera. It also previews on Apple Watch. You can also use the Ecobee thermostat as your doorbell chime. Keep in mind that it's wired only, so it won't work for everyone. Still, the Ecobee Smart Doorbell Camera retails for $160, is available today, and I'd love to use it to replace my current two-camera setup.

3. Samsung S95D OLED TV

Okay, I should preface this by saying I would LIKE to buy this TV as soon as it's released — whether I can afford to buy it will be a different matter. (We'll see when it comes out this spring.) Nevertheless, I've already called the S95D "the best TV of CES 2024" and the TV with the best picture quality in the world right now. Last year's model (S95C) was already among the consensus top three TVs in the world based on picture quality — the other two being the LG G3 and the Sony A95L — at least among TVs consumers can buy at retail.

Also: I saw Samsung's CES 2024 deluge of new TV tech and these 4 products impressed me

And then Samsung announced a new feature at CES 2024 that changed the game. The S95D now has a glare-free screen and in my time with the product at CES I was blown away by how effective it is. Glare is one of the biggest drawbacks on virtually all OLED TVs. They have amazing dynamic range, intense colors, and impressive black levels, but it's hard to use them in bright rooms because they don't have the strong backlight of LCD and QLED TVs and because they become mirrors for windows and lamp lights. The S95D overcomes this with its new anti-reflective display. We tested it with a bright flashlight on a smartphone camera and was amazed at how well it dispersed the intense light and turned it into a soft glow that the TV's picture could overwhelm. That's why I consider it the TV with the best picture quality a consumer can buy, and why I'd like to get one. Pricing won't be announced until spring, but I expect it will start around $2,500 for a 55-inch model before it gets discounted during the fall shopping season.

4. Vasco Translator E1

One of the most impressive demos I had at CES 2024 was with the team of the Vasco Translator E1, an earpiece that uses AI and an app to translate 49 languages in real time. We tested it with a member of the Vasco team speaking a Polish, my ZDNET colleague Sabrina Ortiz speaking in Spanish, and me speaking in English and the E1 earpiece automatically translated between the three languages in our ear as well as displaying the translations in text on the phone app. There was a slight delay and the translations were about 80-90 percent accurate, but it was remarkable how well it let us communicate across the barrier of three different languages. The product will be released in Q2 2024 and pricing isn't available yet.

Also: The 12 best mobile accessories at CES 2024: iPhone call recorder, Qi2 chargers, and more

The team already has a product available, the Vasco Translator V4, a handheld mobile device that translates 108 languages and has a built-in SIM card that works in roughly 200 countries and includes connectivity for the lifetime of the device. The V4 costs $389 and it will work with the E1 earpiece when it's released. My colleagues Kerry Wan and Sabrina Ortiz also demoed a competing product, the TimeKettle X1 Interpreter Hub, and found it to be a little more accurate than Vasco's products and it worked better in loud spaces. However, in the volunteer work I do in community building and with a youth literacy program, I regularly collaborate with immigrants who speak less common languages that are covered by Vasco (such as Swahili and Persian) and are not covered by TimeKettle, and that's why I'm more interested in Vasco's product. Oh, and the fact that the TimeKettle device costs $700 — although it does include two earpieces similar to the E1 as well.

5. Oclean X Ultra S smart toothbrush

Smart toothbrushes have a long and interesting history at CES. I've lost count of how many I've demoed over the years. But none of them ever convinced me to put aside my old school Philips Sonicare toothbrush that I've been using for almost a decade (while regularly replacing the brush heads, of course). This year I think I might have found the one I've been waiting for, though. The Oclean X Ultra S has a built-in display to give you instant feedback on any spots you missed and a brushing score, WiFi and Bluetooth connectivity, a bone-conducting speaker that tells you if you're brushing too fast or applying too much pressure, Apple Health integration, and 4-6 months of battery life.

The Oclean X Ultra S will come to the US market in Q3 and will cost $130. Considering that high-end smart toothbrushes from market leaders Philips and Oral-B cost $300 to $350, the Oclean X Ultra S can bring tech-enabled brushing to a lot more people. The Oclean app is currently not rated nearly as well as Philips, Oral-B, Colgate, and other smart toothbrush apps on the Apple App Store, so Oclean may still have some work to do to improve usability and the overall digital experience. But based on what I saw at CES 2024, this product has a lot going for it and is coming in at a price that I finally feel comfortable paying for a smart toothbrush.

CES 2024

Since Copilot's initial launch, Microsoft's chatbot has proven to be a worthy ChatGPT rival — and now a new wave of updates will close the gap between both chatbots even further.

Microsoft has announced the availability of Copilot Pro, a supercharged subscription version of Copilot, which is meant for power users who could benefit from advanced help with writing, coding, designing, and more.

Also: AI will have a big impact on jobs this year. Here's why that could be good news

The Copilot Pro offerings include a single AI experience across all of your devices to give you better assistance, priority access to the latest models, including OpenAI's GPT-4 Turbo, enhanced AI image creation, and the ability to create your own Copilot GPTs.

If you are a Microsoft 365 Personal and Family subscriber, you'll be especially pleased to know that you can use a Copilot Pro subscription to access Copliot in Word, Excel, PowerPoint, Outlook, and OneNote on PC, Mac, and iPad.

Also: How AI-assisted code development can make your IT job more complicated

The Copilot Pro subscription costs $20 per month per user, which is the same price as ChatGPT's premium subscription, ChatGPT Plus.

Microsoft has also announced that the enterprise offering of Copilot for Microsoft 365 will no longer have a seat minimum, which makes it possible for businesses of all sizes to take advantage of Copilot's assistance.

Also: 6 ways business leaders are exploring generative AI at work

Copilot for Microsoft 365 is available for small businesses with Microsoft 365 Business Premium and Business Standard plans, which can be purchased for between one and 299 seats at $30 per person per month, according to the release.

Other plan updates include the removal of the 300-seat purchase minimum for commercial plans, the availability of Copilot for Office 365 E3 and E5 enterprise customers, and the availability of Copilot for Microsoft 365 for purchase for businesses through a network of Microsoft Cloud Solution Provider partners.

In addition to the Copilot Pro news, the standard Copilot will get a wave of upgrades, starting with Copilot GPTs. Microsoft said a handful of Copliot GPTs will start rolling out today. These GPTs will serve a specific purpose for users, such as fitness, travel, cooking, and more. Only Copilot Pro users, however, will be able to make their own.

Also: The 15 best robots and AI tech we saw at CES 2024

The release of these individual Copilots is similar to the 28 chatbots that Meta announced in September, which have their own personas to serve different purposes. For example, you can chat with Max the chabot, in the style of Roy Choi, for cooking tips and tricks.

The Copilot app is now available for Android and iOS, and a Copilot to the Microsoft 365 mobile app for Andorid and iOS users with a Microsoft account will begin to roll out during the next month.

Artificial Intelligence

Microsoft launches a Pro plan for Copilot Kyle Wiggers 9 hours

Microsoft evidently envisions Copilot, the umbrella brand for its portfolio of AI-powered, content-generating technologies, becoming a significant future revenue line-item. And that’s perhaps not far off base; according to the company, more than 40% of the Fortune 100 participated in its Copilot early access program.

But given the enormous cost of running GenAI models in the cloud, getting Copilot from expenditure to reliable revenue generator will require sustained — and large-scale, ideally — growth.

Surely aware of this, Microsoft is today launching a consumer-focused paid Copilot plan and loosening the eligibility requirements for enterprise-level Copilot offerings. The goal, it appears, is to broaden the base of potential paying Copilot customers while making Microsoft’s existing services — namely Word, Excel and the other apps within the tech giant’s Microsoft 365 family — more attractive through AI features.

Copilot Pro — the new consumer plan, priced at $20 per user per month — gives customers access to Copilot GenAI features across Word, Excel (in preview, only in English for now), PowerPoint, Outlook and OneNote on PC, Mac and iPad — if they have a Microsoft 365 Personal or Family plan, that is. Copilot Pro doesn’t come bundled with a Microsoft 365 subscription. As with the Copilot enterprise offering (Copilot for Microsoft 365), it’s a premium add-on — bringing the total cost of the lowest-tier Microsoft 365 subscription to $27 per month ($6.99 per month for Microsoft 365 Personal plus $20 for Copilot Pro).

The Microsoft 365 capabilities in tow with Copilot Pro are the same that enterprise customers have had for a while.

In Word and OneNote, Copilot writes, edits, summarizes and generates text. Copilot in Excel and PowerPoint turns natural language commands into designed presentations and data visualizations. And in Outlook, Copilot helps drafts email responses with toggles to adapt the length or tone.

Beyond the Microsoft 365 upgrades, Copilot Pro subscribers get 100 “boosts” per day in Designer (formerly Bing Image Creator), Microsoft’s AI-powered image creation tool, to speed up the image generation process — plus improved generation quality and landscape formatting options. And they have priority access to the newest GenAI models underpinning Copilot, including OpenAI’s GPT-4 Turbo, for what Microsoft claims is better performance during peak times.

In the future, Copilot Pro subscribers will be able to switch between models depending on their preferences and, if they require even greater customization, tap Microsoft’s forthcoming Copilot GPT Builder to create “Copilots” tailored for specific topics from sets of prompts.

Copilot GPT Builder sounds suspiciously like OpenAI’s recently released GPT Builder for creating custom chatbots powered by OpenAI GenAI models. But one presumes that Copilot GPT Builder will come with Microsoft service- and app-specific integrations.

Copilot for business

As Microsoft rolls out a premium Copilot for consumers, it’s broadening the service’s business availability, as well.

Starting today, Copilot is generally available for organizations subscribed to Microsoft 365 Business Premium, Microsoft 365 Business Standard, Microsoft 365 E3 and E5 or Office 365 E3 and Office E5. Previously, Copilot for Microsoft 365 had a 300-user minimum purchase and required a Microsoft 365 license, but both of those requirements have been done away with.

There’s a few differences to note between Copilot for Microsoft 365 and Copilot Pro, the main one being Copilot in Teams. Enterprise Copilot customers — not consumers — get a “Copilot” in Teams that provides real-time summaries and action items, handling tasks such as identifying people for follow-ups and creating meeting agendas.

The different Copilot plans, compared.

In addition, Copilot for Microsoft 365 comes with what Microsoft describes as “enterprise-grade data protection” and the Semantic Index, a back-end system that creates a map of the data and content in an organization to allow Copilot to deliver ostensibly more personal and relevant responses.

Copilot for Microsoft 365 customers can also access expanded customization options via Copilot Studio, a souped-up version of Copilot GPT Builder. Unveiled in November, Copilot Studio lets users build their own chatbots and plugins and conduct fine-tuning with first-party company data.

New free features

Microsoft’s attention might be turning toward paid Copilot plans, but the company’s not completely neglecting free users.

Today marks the launch of Copilot GPTs, which like OpenAI’s GPTs are tailored to topics of particular interest. A handful of Copilot GPTs rolled out this morning on the web client for Copilot, fine-tuned to answer questions about things like fitness, travel and cooking.

A free mobile app for Copilot — with access to GPT-4, DALL-E 3 for image creation and the ability to use images on a phone while chatting with Copilot, as well as chat history syncing between mobile, PC and the web — is now live for Android and iOS. And Microsoft says that it’s adding Copilot to the Microsoft 365 mobile app for Android and iOS for users with a Microsoft account. Set to roll out over the coming month, the Microsoft 365 mobile app Copilot integration will let users export content created with Copilot to a Word or PDF document.

Lastly, Microsoft says it’s expanding the number of languages Copilot supports. In the first half of 2024, Copilot will expand to Arabic, Czech, Danish, Dutch, Finnish, Hebrew, Hungarian, Korean, Norwegian, Polish, Portuguese, Russian, Swedish, Thai, Turkish and Ukrainian.

Will startup valuations start to recover in 2024? Investors aren’t so sure Rebecca Szkutak 10 hours

In 2021, it felt like every startup was able to raise at an inflated valuation no matter its size, sector or underlying business model. Today, things look a lot different.

Comparing pre-money valuations, every startup fundraising stage except seed saw median valuations decline last year compared to 2022, according to data from PitchBook. Things were slightly better in 2022, when only the median late-stage and growth-stage valuations were down from 2021, while the median early-stage valuation continued to rise.

Things aren’t looking so good this year either. A recent TechCrunch+ survey of more than 40 investors found that very few VCs actually expect valuations to rise again this year. In fact, a lot of VCs said valuations will continue to drop, while others think we are already at the bottom.

However, they all agreed on one thing: In 2024, stage and sector will matter now more than ever for determining valuation trends.

Early stage

When the market started to turn in 2022, seed and early-stage valuations did not decline as quickly as the late stage, because younger startups are more insulated from the public markets. Because of that delay, some investors think there is still room for seed valuations to come down.

Kirby Winfield, founding general partner at Ascend, predicted that seed valuations will likely keep declining another 5% to 10% before they normalize. Drew Glover, a general partner at Fiat Ventures, also thinks we aren’t at the bottom quite yet.

“At the earliest stages, we’ll continue to see those valuations come back down to earth, but overall, settle in a position that everyone feels like it’ll provide value to investors and to the employees of those companies as well,” Glover said.

The year of ‘does this serve us’ and the rejection of reification

Why we should remember that 'inevitable' never is

Darrell Etherington @etherington / 8 hours

2024 has arrived, and with it, a renewed interest in artificial intelligence, which seems like it’ll probably continue to enjoy at least middling hype throughout the year. Of course, it’s being cheerled by techno-zealot billionaires and the flunkies bunked within their cosy islands of influence, primarily in Silicon Valley – and derided by fabulists who stand to gain from painting the still-fictional artificial general intelligence (AGI) as humanity’s ur-bogeyman for the ages.

Both of these positions are exaggerated and untenable, e/acc vs. decel arguments be damned. Speed without caution only ever results in compounding problems that proponents often suggest are best-solved by pouring on more speed, possibly in a different direction, to arrive at some idealized future state where the problems of the past are obviated by the super powerful Next Big Thing of the future; calls to abandon or regress entire areas of innovation meanwhile ignore the complexity of a globalized world where cats generally can not be put back into boxes universally, among many, many other issues with that kind of approach.

The long, thrilling and tumultuous history of technology development, particularly in the age of the personal computer and the internet, has shown us that in our fervor for something new, we often neglect to stop and ask ‘but is the new thing also something we want or need.’ We never stopped to ask that question with things like Facebook, and they ended up becoming an inextricable part of the fabric of society, an eminently manipulable but likewise essential part of crafting and sharing in community dialog.

Here’s the main takeaway from the rise of social media that we should carry with us into the advent of the age of AI: Just because something is easier or more convenient doesn’t make it preferable — or even desirable.

LLM-based so-called ‘AI’ has already infiltrated our lives in ways that will likely prove impossible to wind back, even if we wanted to do such a thing, but that doesn’t mean we have to indulge in the escalation some see as inevitable, wherein we relentlessly rip out human equivalents of some of the gigs that AI is already good at, or shows promise in, to make way for the necessary ‘forward march of progress.’

The oft-repeated counter to fears that increased automation or handing menial work over to AI agents is that it’ll always leave people more time to focus on ‘quality’ work, as if dropping a couple of hours per day spent on filling in Excel spreadsheets will leave the office admin who was doing that work finally free to compose the great symphony they’ve had locked away within them, or to allow the entry-level graphic designer who had been color-correcting photos the liberty to create a lasting cure for COVID.

In the end, automating menial work might look good on paper, and it might also serve the top executives and deep-pocketed equity-holders behind an organization through improved efficiency and decreased costs, but it doesn’t serve the people who might actually enjoy doing that work, or who at least don’t mind it as part of the overall mix that makes up a work life balanced between more mentally taxing and rewarding creative/strategic exercises and day-to-day low-intensity tasks. And the long-term consequence of having fewer people doing this kind of work is that you’ll have fewer overall who are able to participate meaningfully in the economy — which is ultimately bad even for those rarified few sitting at the top of the pyramid who reap the immediate rewards of AI’s efficiency gains.

Utopian technologist zeal always fails to recognize that the bulk of humanity (techno-zealots included) are sometimes lazy, messy, disorganized, inefficient, error-prone and mostly satisfied with the achievement of comfort and the avoidance of boredom or harm. That might not sound all that aspirational to some, but I say it with a celebratory fervor, since for me all those human qualities are just as laudable as less attainable ones like drive, ambition, wealth and success.

I’m not arguing against halting or even slowing the development of promising new technology, including LLM-based generative AI. And to be clear, where the consequences are clearly beneficial — e.g., developing medical image diagnosis tech that far exceeds the accuracy of trained human reviewers, or developing self-driving car technology that can actually drastically reduce the incidence of car accidents and loss of human life — there is no cogent argument to be made for turning away from use of said tech.

But in almost all cases where the benefits are painted as efficiency gains for tasks that are far from life or death, I’d argue it’s worth a long, hard look at whether we need to bother in the first place; yes, human time is valuable and winning some of that back is great, but assuming that’s always a net positive ignores the complicated nature of being a human being, and how we measure and feel our worth. Saving someone so much time they no longer feel like they’re contributing meaningfully to society isn’t a boon, no matter how eloquently you think you can argue they should then use that time to become a violin virtuoso or learn Japanese.

With large language models (LLMs), a whole new wave of generative AI turned the game upside down in 2023. The new ecosystem – techniques, tools, vendors – has even left AI and machine learning veterans scrambling to wrap their heads around the possibilities and figure out the real deal resources.

While the community continues to experiment and explore LLMs, here are six high-quality courses to learn everything about these language models:

Full Stack LLM Bootcamp

Deep Learning AI conducts the LLM course personally recommended by AI genius Andrej Karpathy. The two-day program is based on the current best practices and the research results to help generative AI developers make the transition to building applications around language models with confidence.

The instructors ran the program as an in-person boot camp in San Francisco in April 2023. Now, they have released the recorded lectures for free.

The course’s objective is to teach regular Python coders how to build applications of LLMs. Instead of the old-school way of building apps with pre-trained models, APIs can be configured and functioning in an hour.

Generative AI with Large Language Models

The free course on Coursera is like AI 101 as it covers the basics, practical stuff, and an in-depth understanding of how these generative AI models actually work. The instructors, Antje Barth, Shelbee Eigenbrode, Mike Chambers, and Chris Fregly, will break down the latest research and show you how companies are actively building LLMs. The course provided by Deep Learning AI and AWS has 166,689 already enrolled for it.

Generative AI Fundamentals

Software leader Databricks has actively contributed to the language model landscape over the past year. Apart from building and open-sourcing its own language models, the company also started dishing out free on-demand knowledge on the ABCs of generative AI.

It’s a four-part video session explaining extensively what you can do with large language models. They’re also making sure you learn about the risks and challenges that come with this tech. The Databricks-style crash course is a great point to start learning about generative AI.

Hugging Face NLP Course

Most of the large language models released in the past year have been hosted on Hugging Face. Even models by big tech companies like Meta’s Llama are available on the developers’ platform. The platform also hosts the Hugging Face Course that would equip you with the right skills to understand the basic concepts of Natural Language Processing (NLP) and language models simultaneously.

Moreover, the course is like a secret stroll through the Hugging Face environment. The course is a backstage pass to the generative AI party.

Introduction to Large Language Models

The course presented by Google Cloud is a quick introductory course that breaks down what LLMs are, where they can be used and how to make them better through prompt tuning. It also covers Google tools to help you develop generative AI applications.

LLM University

Lastly, Cohere’s LLM University offers a deep dive into NLP techniques. The platform has got you covered in everything from semantic search and generation to classification and embeddings. It’s a one-stop shop with a mix of theory and hands-on exercises, giving you knowledge on a bunch of topics. Whether you’re just starting out or a seasoned pro, this platform will polish your LLM skills.

The post 6 Free Courses on Large Language Models appeared first on Analytics India Magazine.