Latest news with #NoPriors
Business Insider
13-06-2025
- Business
- Business Insider
Top AI researchers say language is limiting. Here's the new kind of model they are building instead.
As OpenAI, Anthropic, and Big Tech invest billions in developing state-of-the-art large-language models, a small group of AI researchers is working on the next big thing. Computer scientists like Fei-Fei Li, the Stanford professor famous for inventing ImageNet, and Yann LeCun, Meta's chief AI scientist, are building what they call "world models." Unlike large-language models, which determine outputs based on statistical relationships between the words and phrases in their training data, world models predict events based on the mental constructs that humans make of the world around them. "Language doesn't exist in nature," Li said on a recent episode of Andreessen Horowitz's a16z podcast. "Humans," she said, "not only do we survive, live, and work, but we build civilization beyond language." Computer scientist and MIT professor, Jay Wright Forrester, in his 1971 paper "Counterintuitive Behavior of Social Systems," explained why mental models are crucial to human behavior: Each of us uses models constantly. Every person in private life and in business instinctively uses models for decision making. The mental images in one's head about one's surroundings are models. One's head does not contain real families, businesses, cities, governments, or countries. One uses selected concepts and relationships to represent real systems. A mental image is a model. All decisions are taken on the basis of models. All laws are passed on the basis of models. All executive actions are taken on the basis of models. The question is not to use or ignore models. The question is only a choice among alternative models. If AI is to meet or surpass human intelligence, then the researchers behind it believe it should be able to make mental models, too. Li has been working on this through World Labs, which she cofounded in 2024 with an initial backing of $230 million from venture firms like Andreessen Horowitz, New Enterprise Associates, and Radical Ventures. "We aim to lift AI models from the 2D plane of pixels to full 3D worlds — both virtual and real — endowing them with spatial intelligence as rich as our own," World Labs says on its website. Li said on the No Priors podcast that spatial intelligence is "the ability to understand, reason, interact, and generate 3D worlds," given that the world is fundamentally three-dimensional. Li said she sees applications for world models in creative fields, robotics, or any area that warrants infinite universes. Like Meta, Anduril, and other Silicon Valley heavyweights, that could mean advances in military applications by helping those on the battlefield better perceive their surroundings and anticipate their enemies' next moves. The challenge of building world models is the paucity of sufficient data. In contrast to language, which humans have refined and documented over centuries, spatial intelligence is less developed. "If I ask you to close your eyes right now and draw out or build a 3D model of the environment around you, it's not that easy," she said on the No Priors podcast. "We don't have that much capability to generate extremely complicated models till we get trained." To gather the data necessary for these models, "we require more and more sophisticated data engineering, data acquisition, data processing, and data synthesis," she said. That makes the challenge of building a believable world even greater. At Meta, chief AI scientist Yann LeCun has a small team dedicated to a similar project. The team uses video data to train models and runs simulations that abstract the videos at different levels. "The basic idea is that you don't predict at the pixel level. You train a system to run an abstract representation of the video so that you can make predictions in that abstract representation, and hopefully this representation will eliminate all the details that cannot be predicted," he said at the AI Action Summit in Paris earlier this year. That creates a simpler set of building blocks for mapping out trajectories for how the world will change at a particular time. LeCun, like Li, believes these models are the only way to create truly intelligent AI. "We need AI systems that can learn new tasks really quickly," he said recently at the National University of Singapore. "They need to understand the physical world — not just text and language but the real world — have some level of common sense, and abilities to reason and plan, have persistent memory — all the stuff that we expect from intelligent entities."
CNBC
10-06-2025
- Business
- CNBC
32. Harvey
Founders: Winston Weinberg (CEO), Gabe PereryaLaunched: 2022Headquarters: San FranciscoFunding: $500 millionValuation: $3 billionKey Technologies: Artificial intelligence, generative AIIndustry: Enterprise technology, legal servicesPrevious appearances on Disruptor 50 list: 0 In law firms, young associates are expected to grind through endless hours of research and document review. Harvey, an AI-powered legal technology company, offers an alternative: automation that frees up associates to focus on strategy. The three-year-old startup is part of a new wave of AI companies targeting professional services and reshaping knowledge work. Founded by a former lawyer and a Google AI researcher, Harvey builds large language models tailored for legal and compliance work. Its software handles tasks like analyzing and drafting contracts, due diligence, litigation support, and compliance workflows. It's caught on with some of the world's largest companies and law firms. Harvey has more than 250 enterprise customers in 43 countries, including A&O Shearman, Verizon and PwC. In September 2024, it announced a collaboration with PwC to co-build the Tax AI Assistant, a suite of models trained on curated tax data and refined with feedback from PwC's own experts. The partnership highlights Harvey's strategy of working closely with firms to tailor its models. Despite the fear of AI replacing knowledge workers, Harvey's CEO and co-founder Winston Weinberg, a former lawyer himself, says Harvey aids lawyers but doesn't replace them. "It's not job displacement, it's task displacement," he said on the "No Priors" podcast. "Getting rid of those tasks doesn't mean the legal industry falls apart. It evolves." Investors have taken notice. The company raised $300 million in February in a Series D round that valued the company at $3 billion. Investors included Sequoia, Kleiner Perkins, GV, Elad Gil, Conviction and fellow Disruptor OpenAI's startup fund. The round followed soon after a $100 million fundraising last July, and Reuters recently reported that the company is in talks to raise another $250 million at a higher valuation. Meanwhile, the company continues to add new features. It launched a new version of its Vault storage, search and insights product in November, hired a CFO in March, and opened a London office in September to support growing international demand. Harvey is betting that law firms and professional services companies will continue to embrace AI that trims the busywork and enhances their highest-value output. The billable hour may not be dead, but how those hours are spent is starting to change.
Business Insider
06-06-2025
- Business
- Business Insider
Big Tech is vibe coding with these winning AI startups
It's getting clearer who the winners will be in key parts of the generative AI race, according to Elad Gil, a top startup investor. "In coding, it seems like it's consolidated into 2 or 3 players," he said recently on my favorite AI podcast, " No Priors." He highlighted Cursor, Codium (now called Qodo), Cognition AI (the startup behind Devin), and Microsoft's GitHub Copilot. A clear sign of progress in the tech industry is when a giant platform decides to use an outside service rather than its own product. Google, Amazon, Microsoft, and others have thousands of engineers who can whip up new tech pretty well. So it's a major signal when these companies decide that, no, their home-grown stuff may not be enough. This is happening with Cursor, an AI coding tool from startup Anysphere. Amazon is working on making this available to its employees, according to a scoop this week from Business Insider's Eugene Kim. Amazon already has its own AI coding assistant, Q, and is developing a more advanced tool codenamed " Kiro." So this is a notable move for a company that had warned employees about using third-party AI tools. Google has its own internal AI coding tools, too. And yet, CEO Sundar Pichai said this week he's been messing around with Cursor and a similar service called Replit, building a custom webpage for himself. Software engineering is evolving from a specialized skill into something that non-technical folks can try. Nvidia CEO Jensen Huang likes to say that everyone is a programmer now. Instead of learning complex coding languages, we can create digital things using plain English. Still, some AI coding tools require more expertise than others. Cursor is an IDE, or integrated developer environment, a common setup for pro software engineers. Replit and another coding tool called work in a browser and are considered more user-friendly for novices. Pichai made the distinction this week, saying he uses Cursor, and has "vibe coded with Replit." Vibe coding is a hot new phrase for some of these easier-to-use tools. A good rule of thumb: If you didn't know what IDE stands for, you probably aren't ready for Cursor! Here are more tips.
Yahoo
20-05-2025
- Business
- Yahoo
Duolingo CEO says AI is a better teacher than humans—but schools will still exist ‘because you still need childcare'
Duolingo's founder and CEO Luis von Ahn believes there's nothing a computer can't teach—but says schools won't go extinct because people need childcare. Speaking on the No Priors podcast, von Ahn said AI's precision knowledge and tricks the company has learned about human motivation make a case for 'scaling up' learning in a way that goes beyond humans. Language-learning app Duolingo has been leaning heavily into AI. The company with an owl mascot temporarily replaced its CEO with an AI avatar on an earnings call last year—and evenmore controversially, it announced last month it would permanently replace its contract workers with AI. Now the company has much broader ambitions. With a community of 116 million users a month, Duolingo has amassed loads of data about how people learn, accumulating tricks to keep learners engaged over the long term and even know how well a student will score on a test before they take it. According to founder and CEO Luis von Ahn, AI's ability to individualize learning will lead to most teaching being done by computers in the next few decades. 'Ultimately, I'm not sure that there's anything computers can't really teach you,' von Ahn said on the No Priors podcast recently. He predicted education would radically change, because 'it's just a lot more scalable to teach with AI than with teachers.' 'By the way, that doesn't mean the teachers are going to go away, you still need people to take care of the students,' he added. 'I also don't think schools are going to go away, because you still need childcare.' Host Sarah Guo jumped in to clarify. 'In your view, schools could be childcare but everybody's Duolingo-ing?' she said. 'I think it's going to be something like that,' von Ahn replied. The Duolingo model of teaching via quizzes and drills isn't suitable for all subjects, he said, noting that history might be a subject better taught with 'well-produced videos'—something AI can't currently do well. But he said he believes the problem of scale tips the balance on the side of AI. If 'it's one teacher and like 30 students, each teacher cannot give individualized attention to each student,' he said. 'But the computer can. And really, the computer can actually … have very precise knowledge about what you, what this one student is good at and bad at.' Duolingo's CFO made similar comments last year, saying, 'AI helps us replicate what a good teacher does'—things like helping a student 'learn material, stay engaged, know where your weaknesses are, where your gaps are.' Because Duolingo has acquired this data over many years and millions of users, the company has essentially run 16,000 A/B tests over its existence, von Ahn said. That means the app can deploy reminders at the time that a person is most likely to do a task, and devise exercises that are exactly the right amount of difficulty to keep students feeling accomplished and moving ahead. Some schools are already leaning into AI. Newsweek recently profiled Alpha School, a chain of private K-12 schools where students learn for just two hours a day with the help of AI. There, guides—the school's title for teachers—'provide motivational and emotional support rather than creating lesson plans, delivering lectures or grading assignments,' Newsweek wrote. With four locations and eight on the way, the school charges $40,000 to $65,000 a year in tuition, according to its website. With President Donald Trump recently signing an executive order to promote AI education, manymore schools could be seeing the technology. Von Ahn, at least, thinks the change won't happen for some years. 'I don't think you'll see a change where next year everybody's learning is completely different,' he said. When it comes to most education, 'it's like government—it's just slow.' This story was originally featured on
Time of India
20-05-2025
- Business
- Time of India
Duolingo CEO Luis von Ahn makes a prediction on how AI will change schools in future
Duolingo CEO Luis von Ann recently predicted that artificial intelligence (AI) will fundamentally transform education, shifting the role of schools from traditional learning institutions to childcare and supervised spaces. As reported by Business Insider, speaking at the No Priors podcast, von Ahn stressed on the fact that AI's scalability makes it more efficient than human teachers and it can be easily used for personalised education. "Education is going to change. It's just a lot more scalable to teach with AI than with teachers,' von Ahn said. Duolingo CEO Luis von Ahn feels that AI will take over teaching Von Ahn suggests that AI will increasingly handle the core instruction in schools. He believes that AI can deliver personalised education more effectively than traditional classroom models with teachers. AI system can be designed to precisely track individual student progress, identify areas where they struggle and then tailor the learning experience accordingly. Schools won't disappear, but their role will change: Duolingo CEO Luis von Ahn While Von Ahn does not believe schools will vanish, he foresees a future where AI tutors handle most of the teaching, while schools primarily serve as safe environments for children. He explained that AI can track individual student performance in real time, adjusting lesson plans to match each child's learning pace—something human teachers struggle to do in large classrooms. 'That doesn't mean the teachers are going to go away. You still need people to take care of the students. I also don't think schools are going to go away because you still need childcare,' von Ahn said on the podcast. This doesn't mean teachers will become obsolete, according to von Ahn. Instead, their responsibilities may evolve. Teachers could transition from primarily delivering instruction to supervising students, providing guidance, and fostering social and emotional development. Von Ahn also points out that the integration of AI in education is already happening. Duolingo, for example, is restructuring its operations to embrace an "AI-first" approach. This includes using AI for tasks previously done by contractors and incorporating AI into employee performance reviews. AI Masterclass for Students. Upskill Young Ones Today!– Join Now
