Рубрика: AI

Gemini on Android can’t ID songs, and it’s frustrating Kyle Wiggers 8 hours

If it wasn’t clear before that Google’s Gemini chatbot was rushed out the door, it is now.

Gemini’s since-removed image generator put people of color in Nazi-era uniforms. The chatbot’s commentary continues to tend toward the absurd besides, like equating Hitler’s record with Elon Musk posting memes.

On Android, Gemini also breaks Google Assistant’s song recognition. And to me, after Gemini’s abhorrent cultural insensitivities, it’s one of the most frustrating things about it.

Let me explain.

I mostly use Gemini on my aging Samsung Galaxy A53 5G, which isn’t exactly the zippiest Android smartphone out there. To make it snappier, I replaced the default home screen with a minimalist alternative, Niagara Launcher, which is essentially an alphabetized list of the apps installed on my phone.

Niagara’s great. But it’s limited in what it can do by design, which made me dependent on Google Assistant — now Gemini — for tasks like setting timers, launching apps and so on.

Song recognition, prompted with a command like “OK Google, what’s thing song?,” was a Google Assistant-delivered convenience I took advantage of often. It came in handy in nightclubs, restaurants and bars for IDing tracks I’d most certainly forget otherwise. There’s no shortage of song-recognizing apps — Apple-owned Shazam to name one. But Google Assistant was among the better out there in terms of accuracy, at least in my experience.

So imagine my frustration when I discovered that Gemini on Android can’t recognize songs — or even perform the basic task of funneling song ID requests to Google Assistant.

Ask Gemini, which replaces Google Assistant on Android, to ID a song and it has the nerve to suggest using apps like Shazam — or invoking Google Assistant by switching back to it. For added variety, it’ll occasionally suggest random songs from YouTube.

Image Credits: Google

I’m acutely aware this is a first-world problem. I could launch song recognition from the Google Search app on Android. Or, were I using a conventional home screen, I could place the dedicated song ID shortcut.

But the other aspect of Google Assistant’s song recognition that made it so attractive, at least to this writer, was the low barrier to use. Launching the feature didn’t require fiddling with an app or typing anything in. A voice command later and it was up and running, which made it fast — useful when you’re trying to quickly ID a song.

Making the song recognition situation more disheartening is the fact that I’m paying for the Google One AI Premium Plan, priced at $20 per month, which is supposed to afford me access to a more sophisticated, powerful Gemini experience. Perhaps it’s sophisticated in other ways — ways I haven’t discovered yet frankly. But broken song recognition, along with missing basic features like the ability to play songs, create lists and more, make Gemini a very poor substitute for Google Assistant on Android at present.

Full transparency, I’ve reached out to Google about song recognition via Gemini and I’ll update this post if I hear back.

Salesforce today announced the public beta availability of Einstein Copilot, a new customisable, conversational, and generative AI assistant for CRM. Einstein Copilot is available in beta globally for Sales Cloud and Service Cloud, with Commerce Cloud and Marketing Cloud available later in 2024.

“Unlike other AI assistants or copilots that lack adequate company data to generate useful responses, Einstein Copilot enables Salesforce customers to generate responses using their own private and trusted data, while maintaining strict data governance and without requiring expensive AI model training,” the company said in a press release.

Einstein Copilot can answer questions, summarise content, create new content, interpret complex conversations, and dynamically automate tasks on behalf of a user, all from a single, consistent user experience embedded directly within Salesforce’s AI CRM applications.

Customers can tap into the full power of Salesforce with Einstein 1 Editions, simplified technology packages for businesses looking to accelerate growth and speed productivity with the AI CRM.

Einstein 1 Editions provide organizations in every industry access to the best of Salesforce technology, including CRM, Einstein Copilot, Data Cloud, Slack and Tableau in a single offering, helping them transform their business and deliver stronger customer experiences.

This is made possible by combining a conversational UI, a foundational large language model, and trusted company data that enables Salesforce users to tap into the power of generative AI and interact with their applications in entirely new ways.

“Our new Einstein Copilot brings together an amazing intuitive interface for interacting with AI, world-class AI models and above all deep integration of the data and metadata needed to benefit from AI. Einstein Copilot is the only copilot with the ability to truly understand what is going on with your customer relationships,” said Marc Benioff, Chair & CEO, at Salesforce.

The company said Einstein Copilot grounds its responses with trusted business data from Data Cloud to provide the necessary context for the highest quality outputs. This allows Einstein Copilot to generate more precise and tailored responses based on trusted company data.

Moreover, Einstein Copilot can be customised to accomplish specific sales, service, marketing, commerce, and IT tasks, ensuring company and industry policies are applied. Copilot Builder can create custom actions for Einstein Copilot, Prompt Builder activates custom prompts in the flow of work, and Model Builder uses proprietary AI models to power custom Einstein Copilot functionality.

Currently, Einstein Copilot supports data residency in the US and the English language.

The post Salesforce Announces the Public Beta Availability of Einstein Copilot appeared first on Analytics India Magazine.

Soon after ChatGPT became an internet sensation, a comparable development was underway at Tesla’s Palo Alto headquarters in December 2022. Dhaval Shroff, an engineer working on the company’s autopilot system, pitched a concept to CEO Elon Musk. Shroff proposed a system similar to ChatGPT but tailored for automobiles.

Instead of relying on predefined rules to determine the car’s optimal path, they aimed at using a neural network that learns from extensive training data. This data consisted of millions of examples of human driving behaviour, explained Shroff, a seasoned member of the Tesla team with a decade of experience.

Eight months later, Musk experienced an improvement in the performance of a Full Self-Driving (FSD) vehicle compared to the hundreds he had driven earlier. The smoothness and reliability were attributed to the new version, FSD 12, which introduced the new concept.

Musk believed that this innovation had the potential to not only transform autonomous vehicles, but also to represent a leap toward artificial general intelligence capable of operating in real-world scenarios.

Instead of traditionally relying on hundreds of thousands of lines of code, the new system proposed by Shroff learned to drive by processing billions of video frames depicting human driving behaviour. This approach mirrored the self-training method employed by new LLM chatbots, which generate responses by processing billions of words from human text.

Fully Accelerated

Tesla is not the only company employing an end-to-end, there is also Comma.ai with OpenPilot, which Bengaluru boy Mankaran Singh used to power his Alto through an old Android phone. The news of his FSD journey in India has attracted attention since auto manufacturers often tell us how much computing power is on board to make it happen.

Even Wayve.ai, ventured into some of the toughest streets of London to test its self-driving skills. The team broke some impressive ground. Eight months ago, they released a 9-billion parameter world model that uses video, text, and action inputs to train the systems for on-road behaviours.

In May 2022, Wayve collaborated with Microsoft to leverage Azure, the tech giant’s cloud-based supercomputer, for training its neural network.

Musk has pointed out a consequential aspect of the end-to-end approach: vehicles no longer receive explicit instructions such as “stop at a red light” or “verify before changing lanes”. Instead, it autonomously discerns these actions by “imitating” behaviours observed in the 10 million videos used during training.

This means they’ve been using a dataset of millions of videos and have assessed the drivers on each of these. The machine learning model has been trained to mimic the behaviours of what were deemed as “good drivers”.

In theory, this holds huge potential since the models can generalise more effectively when facing unfamiliar scenarios. Essentially, the model can identify the most appropriate behaviour based on its training rather than getting stuck in predefined instructions.

Hit a Brake

One problem, however, is yet to be overcome. Human drivers, even the most skilled ones, often bend traffic rules. For instance, over 95% of humans tend to roll slowly through stop signs rather than coming to a complete halt.

And since the new FSD system is intentionally designed to imitate human behaviour, the head of the National Highway Safety Board is currently investigating if this behaviour could be deemed acceptable for self-driving cars.

Moreover, despite a decade-and-a-half of reckless spending and extensive road testing, driverless technology is stuck in the pilot phase. “We are seeing extraordinary amounts of spending to get very limited results,” noted Alex Kendall, founder and CEO of Wayve.

This has prompted UK-based firms like Wayve and startups such as Waabi and Ghost to focus heavily on neural networks. Branded as AV2.0, they are optimistic that more competent and cost-effective technology will let them surpass current market leaders.

Self-driving cars have made headlines all these years for various high-profile errors that were hard to overlook. Investors have put in over $100 billion into developing autonomous vehicles, amounting to a third of the cost NASA incurred to put humans on the Moon. As of now, one giant leap for humankind is less expensive than a vehicle that can drive itself.

The post Why Did Tesla Build a ChatGPT for Vehicles? appeared first on Analytics India Magazine.

Google has initiated a private program for a select group of independent publishers, granting them beta access to an unreleased generative AI platform in exchange for feedback and analytics, as per documents reviewed by ADWEEK. This program is part of the Google News Initiative, launched in 2018.

Under the agreement’s terms, publishers must use the tools to generate a specific volume of content over a 12-month period. In return, these news organizations receive a monthly stipend totalling a five-figure sum annually.

Addressing speculations, a Google representative clarified, “This speculation about this tool being used to republish other outlets’ work is inaccurate.” The representative added that the experimental tool is designed responsibly to assist local publishers in producing high-quality journalism using factual content from public data sources, such as a local government’s public information office or health authority. It was emphasized that these tools are not intended to and cannot replace journalists’ essential role in reporting, creating, and fact-checking their articles.

The beta tools let publishers create aggregated content by indexing recently published reports from various organizations, including government agencies and neighbouring news outlets. The process involves summarizing and republishing this information as a new article.

In an October edition of the Local Independent Online News newsletter, Google initially called for news organizations to apply to test these tools. Publisher onboarding for the Google News Initiative (GNI) began in January, and the yearlong program officially commenced in February.

Simultaneously, Google launched the second edition of GNI in India. Despite the company’s public investment in journalism initiatives, there is an ironic concern that Google, with its dominant market position, is constraining journalism. This is evident in Google’s abuse of its market power to compel news publishers to use their content in the redesigned Google News app—a mobile news aggregator heavily focused on Accelerated Mobile Pages (AMP).

With the introduction of generative AI tools, there is a potential concern that the articles produced could divert traffic away from the sources, adversely impacting their businesses. This process is reminiscent of the ripping technique currently employed at Reach plc, but the text is sourced from external outlets in this case.

The mission of GNI has raised questions, particularly as it attempts to address a problem that Google itself has sparked in the first place.

The post Google Gives Publishers Beta Access to AI in Exchange for Content appeared first on Analytics India Magazine.

Diffusion transformers are the key behind OpenAI’s Sora — and they’re set to upend GenAI Kyle Wiggers 13 hours

OpenAI’s Sora, which can generate videos and interactive 3D environments on the fly, is a remarkable demonstration of the cutting edge in GenAI — a bona fide milestone.

But curiously, one of the innovations that led to it, an AI model architecture colloquially known as the diffusion transformer, arrived on the AI research scene years ago.

The diffusion transformer, which also powers AI startup Stability AI’s newest image generator, Stable Diffusion 3.0, appears poised to transform the GenAI field by enabling GenAI models to scale up beyond what was previously possible.

Saining Xie, a computer science professor at NYU, began the research project that spawned the diffusion transformer in June 2022. With William Peebles, his mentee while Peebles was interning at Meta’s AI research lab and now the co-lead of Sora at OpenAI, Xie combined two concepts in machine learning — diffusion and the transformer — to create the diffusion transformer.

Most modern AI-powered media generators, including OpenAI’s DALL-E 3, rely on a process called diffusion to output images, videos, speech, music, 3D meshes, artwork and more.

It’s not the most intuitive idea, but basically, noise is slowly added to a piece of media — say an image — until it’s unrecognizable. This is repeated to build a dataset of noisy media. When a diffusion model trains on this, it learns how to gradually subtract the noise, moving closer, step by step, to a target output piece of media (e.g. a new image).

Diffusion models typically have a “backbone,” or engine of sorts, called a U-Net. The U-Net backbone learns to estimate the noise to be removed — and does so well. But U-Nets are complex, with specially designed modules that can dramatically slow the diffusion pipeline.

Fortunately, transformers can replace U-Nets — and deliver an efficiency and performance boost in the process.

A Sora-generated video. Image Credits: OpenAI

Transformers are the architecture of choice for complex reasoning tasks, powering models like GPT-4, Gemini and ChatGPT. They have several unique characteristics, but by far transformers’ defining feature is their “attention mechanism.” For every piece of input data (in the case of diffusion, image noise), transformers weigh the relevance of every other input (other noise in an image) and draw from them to generate the output (an estimate of the image noise).

Not only does the attention mechanism make transformers simpler than other model architectures but it makes the architecture parallelizable. In other words, larger and larger transformer models can be trained with significant but not unattainable increases in compute.

“What transformers contribute to the diffusion process is akin to an engine upgrade,” Xie told TechCrunch in an email interview. “The introduction of transformers … marks a significant leap in scalability and effectiveness. This is particularly evident in models like Sora, which benefit from training on vast volumes of video data and leverage extensive model parameters to showcase the transformative potential of transformers when applied at scale.”

Generated by Stable Diffusion 3. Image Credits: Stability AI

So, given the idea for diffusion transformers has been around a while, why did it take years before projects like Sora and Stable Diffusion began leveraging them? Xie thinks the importance of having a scalable backbone model didn’t come to light until relatively recently.

“The Sora team really went above and beyond to show how much more you can do with this approach on a big scale,” he said. “They’ve pretty much made it clear that U-Nets are out and transformers are in for diffusion models from now on.”

Diffusion transformers should be a simple swap-in for existing diffusion models, Xie says — whether the models generate images, videos, audio or some other form of media. The current process of training diffusion transformers potentially introduces some inefficiencies and performance loss, but Xie believes this can be addressed over the long horizon.

“The main takeaway is pretty straightforward: forget U-Nets and switch to transformers, because they’re faster, work better and are more scalable,” he said. “I’m interested in integrating the domains of content understanding and creation within the framework of diffusion transformers. At the moment, these are like two different worlds — one for understanding and another for creating. I envision a future where these aspects are integrated, and I believe that achieving this integration requires the standardization of underlying architectures, with transformers being an ideal candidate for this purpose.”

If Sora and Stable Diffusion 3.0 are a preview of what to expect with diffusion transformers, I’d say we’re in for a wild ride.

Rising sea levels present a formidable challenge in the fight against climate change, threatening coastal communities, ecosystems and global economies. The urgency to act is paramount, as these waters rise due to melting ice caps and thermal expansion from warming oceans. AI provides hope in this concerning scenario.

This technology’s advanced predictive capabilities and data analysis offer unparalleled opportunities for early detection and mitigation strategies. By harnessing AI, people can forecast future impacts more accurately and develop innovative solutions to protect vulnerable coastlines.

The Rising Challenge of Sea Levels

Two factors linked to global warming primarily drive rising sea levels — melting glaciers and ice sheets, and seawater expanding as it warms. This phenomenon poses a significant threat to coastal cities, risking increased flooding, habitat loss and economic damage. The global average sea level in 2022 stood at 101.2 millimeters, marking a rise of four inches above the levels recorded in 1993. This stark increase underscores the rapid pace at which the planet’s waters rise, highlighting the immediate need for action.

The impact on coastal cities is profound, with communities facing the prospect of displacement, infrastructure damage and compromised water supplies due to saltwater intrusion. These challenges call for innovative solutions that go beyond traditional flood defenses.

Adopting advanced technologies and new strategies develops resilient urban landscapes capable of withstanding the rising tides. The urgency and scale of this issue demand a forward-thinking, adaptable approach to protect coastal areas and ensure their sustainability for future generations.

Understanding AI’s Role in Climate Science

Artificial intelligence is the simulation of human intelligence in machines, enabling them to think and learn. Its relevance in environmental science is monumental because it offers the ability to sift through and analyze vast, complex data sets far beyond human capability.

The technology is crucial for understanding and predicting climate patterns and sea level trends. In addition, integrating data from diverse sources like satellite imagery, ocean temperature readings and atmospheric data is pivotal in this approach.

These models employ sophisticated algorithms to identify patterns and anomalies in climate data for predictions on future climate conditions and rising sea levels. They can simulate various scenarios under different global warming pathways, helping scientists and policymakers devise effective strategies to mitigate climate change impacts.

Moreover, it’s making significant strides in environmental conservation, notably in reducing emissions and addressing plastic waste. These applications underscore AI’s potential as a powerful tool to combat environmental challenges.

AI in Predicting Rising Sea Levels

AI algorithms forecast sea level changes by leveraging data from satellites and ocean sensors. These devices provide continuous information on ocean temperatures, ice sheet masses and sea surface heights. The algorithms analyze this data, identifying patterns and trends that might elude human analysts.

For example, machine learning models can predict the rate at which ice sheets are melting or how much the sea level could rise in specific regions based on current warming trends. This predictive capability is crucial for urban planning and resilience strategies, considering 56% of the world’s population resides in cities. Many of these metropolitan areas are coastal, making them particularly vulnerable to the impacts of rising sea levels.

The improved accuracy of these AI-enhanced climate models helps identify the key drivers of climate change. Scientists and policymakers can pinpoint the most effective intervention points by understanding which factors most significantly affect sea level rise.

For instance, AI models reveal specific emissions disproportionately impact temperature increases so stakeholders can concentrate efforts on reducing those emissions. Similarly, algorithms can help identify regions where intervention could significantly mitigate sea level rise. This guides resource allocation and the development of targeted climate policies.

AI Innovations for Sea Level Mitigation

AI significantly reduces greenhouse gas emissions by optimizing energy use across various industries. Through the deployment of smart algorithms, these systems can analyze patterns in energy consumption and identify opportunities for efficiency improvements.

For instance, AI can adjust heating, ventilation and air conditioning systems in buildings and manufacturing processes in real time. This approach reduces energy use and lowers carbon footprints in the long term. Similarly, it optimizes the supply and demand balance for electricity in the energy sector. It increases the efficiency of renewable energy sources to decrease reliance on fossil fuels.

In carbon capture and storage (CCS) technologies, AI-driven innovations are making strides in reducing atmospheric CO2 levels. Algorithms enhance the efficiency of capture processes by optimizing operational parameters such as temperature, pressure, flow rates and chemical reactions.

This optimization ensures carbon capture facilities operate at peak efficiency, capturing the maximum amount of CO2 with the lowest energy input. Moreover, AI can predict maintenance needs and process adjustments, reducing downtime and improving the overall carbon capture rate.

AI’s predictive capabilities extend to identifying optimal locations for CO2 storage. It analyzes geological data to ensure carbon's safe and permanent sequestration. By improving the capture process and storage capabilities, AI is at the forefront of advancing CCS technologies, which is crucial in the global effort to mitigate climate change.

Preventing the Impact of Rising Sea Levels With AI

AI significantly enhances people’s ability to plan and implement preventive measures against environmental threats like rising sea levels and flooding. Analyzing vast amounts of data can predict when and where adverse events are likely to occur, allowing timely and targeted interventions.

One of the most innovative applications of AI in this context is in the development of automated flood barriers. These systems use AI to monitor real-time data on weather patterns and water levels, automatically activating barriers or floodgates when they detect a threat. Doing so reduces human error and ensures a rapid response to emerging threats, minimizing potential damage.

Similarly, smart drainage systems represent another AI-driven solution. These can optimize water flow in urban areas based on predictive models of rainfall and flooding. By intelligently directing water away from vulnerable areas, smart drainage systems prevent accumulation and reduce the risk of urban flooding. Additionally, they can adapt to changing conditions in real time, ensuring the most effective response to any situation.

AI-Driven Solutions for Coastal Defense

AI-based coastal defense systems integrate data from satellites, weather stations and ocean sensors. They employ sophisticated machine learning algorithms to predict when and where coastal areas are most vulnerable to flooding.

This proactive approach allows adaptive and responsive defense mechanism designs. For instance, predictive models can inform the deployment of flood barriers and optimize drainage systems in real time, ensuring communities are safer against the unpredictable nature of climate change.

Implementing these AI-driven solutions involves a complex network of sensors and data analytics platforms that work together to monitor environmental conditions continuously. This setup enables the system to analyze trends and predict potential flooding events accurately.

When it detects a threat, automated systems can activate defenses or issue warnings to local authorities and residents, providing them with critical lead time to enact emergency plans. The ability to rapidly respond to such threats minimizes physical damage and saves lives by ensuring communities are adequately prepared.

One compelling example of AI’s impact on coastal defense is in Mozambique, where AI-powered models predict flooding patterns and alert communities about impending disasters. These models leverage environmental data to forecast flood risks accurately, allowing timely evacuations and preparations.

Collaborative Efforts and Global Initiatives

Global cooperation in AI research and climate action tackles the multifaceted challenges of climate change. By sharing knowledge, resources and technologies, countries can develop more effective and inclusive solutions that benefit the planet.

Collective efforts enhance the accuracy and applicability of AI models, as they can train on diverse global data sets to ensure the solutions are adequate across different geographical and environmental contexts. This approach accelerates innovation and fosters equity in climate action, allowing developed and developing countries to implement advanced climate mitigation and adaptation technologies.

Investing in AI for a Sustainable Future

Investing in AI research for environmental protection is a necessity for the future. The innovative solutions it offers for predicting, mitigating and adapting to environmental challenges can revolutionize the approach to safeguarding the Earth. By prioritizing and increasing funding for AI technologies, people can unlock unparalleled opportunities to protect ecosystems, combat climate change and ensure a sustainable world for future generations.

IBM announced that LTIMIndtree, a global technology consulting and digital solutions company, has joined the IBM Quantum Network to explore quantum computing innovation for the benefit of its global clientele across multiple industries.

LTIMindtree is the first Indian Global System Integrator (GSI) to join the IBM Quantum Network.

As part of the IBM Quantum Network, LTIMindtree expands on its platinum partner status with IBM and joins a global community of Fortune 500 companies, top universities, research labs, and startups.

LTIMindtree will have access to IBM resources, including IBM’s global fleet of quantum computing systems over the cloud, software, and associated expertise. This move is a strategic step toward LTIMindtree helping their customers benefit from the transformative value of quantum computing technologies.

LTIMindtree will also collaborate with the Indian Institute of Technology (IIT) Madras, which is also an IBM Quantum Innovation Center, on joint quantum research and workforce development.

“These collaborations are more than an innovation milestone; they’re an important step towards a future where quantum computing could help solve more complex problems faster and more efficiently.

“Importantly, it positions us to expedite our customers’ journey towards realizing the immense value of quantum computing, readying them to leverage these advanced technologies for transformative solutions,” said Aan Chauhan, Chief Technology Officer, LTIMindtree.

LTIMindtree’s plans across these collaborations are to establish a series of long-term projects, including applied research toward business and societal problems, quantum computing workshops, and research grants.

These initiatives aim to nurture a new generation of quantum computing professionals and researchers at LTIMindtree, creating a sustainable and innovative ecosystem.

The post LTIMindtree Joins IBM Network to Advance the Quantum Innovation Ecosystem appeared first on Analytics India Magazine.