South Korean President Yoon Suk Yeol has reversed his declaration of martial law after six hours of heightened tension at South Korea’s National Assembly.
Grayscale joins rivals 21Shares, Canary Capital, VanEck and Bitwise in filing for a Solana ETF in the US.
A startup called Exa is pitching a new spin on generative search. It uses the tech behind large language models to return lists of results that it claims are more on point than those from its rivals, including Google and OpenAI. The aim is to turn the internet’s chaotic tangle of web pages into a kind of directory, with results that are specific and precise.
Exa already provides its search engine as a back-end service to companies that want to build their own applications on top of it. Today it is launching the first consumer version of that search engine, called Websets.
“The web is a collection of data, but it’s a mess,” says Exa cofounder and CEO Will Bryk. “There’s a Joe Rogan video over here, an Atlantic article over there. There’s no organization. But the dream is for the web to feel like a database.”
Websets is aimed at power users who need to look for things that other search engines aren’t great at finding, such as types of people or companies. Ask it for “startups making futuristic hardware” and you get a list of specific companies hundreds long rather than hit-or-miss links to web pages that mention those terms. Google can’t do that, says Bryk: “There’s a lot of valuable use cases for investors or recruiters or really anyone who wants any sort of data set from the web.”
Things have moved fast since MIT Technology Review broke the news in 2021 that Google researchers were exploring the use of large language models in a new kind of search engine. The idea soon attracted fierce critics. But tech companies took little notice. Three years on, giants like Google and Microsoft jostle with a raft of buzzy newcomers like Perplexity and OpenAI, which launched ChatGPT Search in October, for a piece of this hot new trend.
Exa isn’t (yet) trying to out-do any of those companies. Instead, it’s proposing something new. Most other search firms wrap large language models around existing search engines, using the models to analyze a user’s query and then summarize the results. But the search engines themselves haven’t changed much. Perplexity still directs its queries to Google Search or Bing, for example. Think of today’s AI search engines as a sandwich with fresh bread but stale filling.
More than keywords
Exa provides users with familiar lists of links but uses the tech behind large language models to reinvent how search itself is done. Here’s the basic idea: Google works by crawling the web and building a vast index of keywords that then get matched to users’ queries. Exa crawls the web and encodes the contents of web pages into a format known as embeddings, which can be processed by large language models.
Embeddings turn words into numbers in such a way that words with similar meanings become numbers with similar values. In effect, this lets Exa capture the meaning of text on web pages, not just the keywords.
Large language models use embeddings to predict the next words in a sentence. Exa’s search engine predicts the next link. Type “startups making futuristic hardware” and the model will come up with (real) links that might follow that phrase.
Exa’s approach comes at cost, however. Encoding pages rather than indexing keywords is slow and expensive. Exa has encoded some billion web pages, says Bryk. That’s tiny next to Google, which has indexed around a trillion. But Bryk doesn’t see this as a problem: “You don’t have to embed the whole web to be useful,” he says. (Fun fact: “exa” means a 1 followed by 18 0s and “googol” means a 1 followed by 100 0s.)
Websets is very slow at returning results. A search can sometimes take several minutes. But Bryk claims it’s worth it. “A lot of our customers started to ask for, like, thousands of results, or tens of thousands,” he says. “And they were okay with going to get a cup of coffee and coming back to a huge list.”
“I find Exa most useful when I don’t know exactly what I’m looking for,” says Andrew Gao, a computer science student at Stanford Univesrsity who has used the search engine. “For instance, the query ‘an interesting blog post on LLMs in finance’ works better on Exa than Perplexity.” But they’re good at different things, he says: “I use both for different purposes.”
“I think embeddings are a great way to represent entities like real-world people, places, and things,” says Mike Tung, CEO of Diffbot, a company using knowledge graphs to build yet another kind of search engine. But he notes that you lose a lot of information if you try to embed whole sentences or pages of text: “Representing War and Peace as a single embedding would lose nearly all of the specific events that happened in that story, leaving just a general sense of its genre and period.”
Bryk acknowledges that Exa is a work in progress. He points to other limitations, too. Exa is not as good as rival search engines if you just want to look up a single piece of information, such as the name of Taylor Swift’s boyfriend or who Will Bryk is: “It’ll give a lot of Polish-sounding people, because my last name is Polish and embeddings are bad at matching exact keywords,” he says.
For now Exa gets around this by throwing keywords back into the mix when they’re needed. But Bryk is bullish: “We’re covering up the gaps in the embedding method until the embedding method gets so good that we don’t need to cover up the gaps.”
This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
Nominate someone to our 2025 list of Innovators Under 35
Every year, MIT Technology Review recognizes 35 young innovators who are doing pioneering work across a range of technical fields including biotechnology, materials science, artificial intelligence, computing, and more.
Previous winners include Lisu Su, now CEO of AMD, Andrew Ng, a computer scientist and serial entrepreneur, Jack Dorsey (two years after he launched Twitter), and Helen Greiner, co-founder of iRobot.
We’re now taking nominations for our 2025 list and you can submit one here. The process takes just a few minutes. Nominations will close at 11:59 PM ET on January 20, 2025. You can nominate yourself or someone you know, based anywhere in the world. The only rule is that the nominee must be under the age of 35 on October 1, 2025. Read more about what we’re looking for here.
How US AI policy might change under Trump
President Biden first witnessed the capabilities of ChatGPT in 2022 during a demo from Arati Prabhakar, the Director of the White House Office of Science and Technology Policy, in the oval office.
That demo set a slew of events into motion, and encouraged President Biden to support the US’s AI sector, while managing the safety risks that will come from it.
However, that approach could change under Trump. Our AI reporter James O’Donnell sat down with Prabhakar earlier this month to discuss what might be next. Read the full story.
This story is from Algorithm, our weekly AI newsletter. Sign up to receive it in your inbox every Monday.
The must-reads
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
1 What’s next for Intel?
Its CEO has been given the boot, and his replacement will be tasked with turning things around. (WSJ $)
+ The departed Pat Gelsinger was firmly opposed to breaking the firm up. (Bloomberg $)
+ Five years ago, Intel was on top of the world. What happened? (FT $)
2 China has hit back at the latest US chip export restrictions
By banning shipments of critical chip minerals to America. (FT $)
+ Beijing accused the US of hindering normal trade exchanges. (The Guardian)
+ What’s next in chips. (MIT Technology Review)
3 Hackers are using AI to mine troves of personal data
The new tools make it much easier to weaponize sensitive information. (WP $)
+ The US government is trying to crack down on the sale of civilians’ personal data. (404 Media)
+ Five ways criminals are using AI. (MIT Technology Review)
4 Elon Musk has been denied a $56 million pay package for a second time
Still, he’s not exactly short of a few bob. (The Verge)
5 A network of women were duped into donating eggs to a disgraced billionaire
The US fertility industry’s loose regulations have left the system open to abuse. (Bloomberg $)
+ Conservative politicians are spreading anti-contraceptive disinformation. (New Yorker $)
+ I took an international trip with my frozen eggs to learn about the fertility industry. (MIT Technology Review)
6 An AI agent could do your next Black Friday shop for you
It could spell an end to tedious price-checking and bargain monitoring. (TechCrunch)
+ What are AI agents? (MIT Technology Review)
7 A new fleet of US nuclear reactors is on the horizon
Similar major pushes have failed in the past. Will this time be different? (The Atlantic $)
+ Why the lifetime of nuclear plants is getting longer. (MIT Technology Review)
8 It turns out that fish have a brain microbiome
It raises the question whether humans could have one too. (Quanta Magazine)
+ The hunter-gatherer groups at the heart of a microbiome gold rush. (MIT Technology Review)
9 Why ChatGPT has become an emotional crutch for so many people
But beware using it to offload emotional labor. It’s only a chatbot, after all. (The Guardian)
+ The name ‘David Mayer’ causes ChatGPT to melt down, for some reason. (TechCrunch)
+ Here’s how people are actually using AI. (MIT Technology Review)
10 This Indigenous community may become Canada’s first climate refugees
The Western Arctic region’s permafrost is thawing, and the Inuvialuit will be forced to leave their homes. (NYT $)
Quote of the day
“It was a tough situation when Pat showed up, and things look much worse now.”
—Financial analysts from Bernstein warn investors that whoever takes over from departing Intel CEO Pat Gelsinger has their work cut out for them, Insider reports.
The big story
How to measure all the world’s fresh water
December 2021
The Congo River is the world’s second-largest river system after the Amazon. More than 75 million people depend on it for food and water, as do thousands of species of plants and animals. The massive tropical rainforest sprawled across its middle helps regulate the entire Earth’s climate system, but the amount of water in it is something of a mystery.
Scientists rely on monitoring stations to track the river, but what was once a network of some 400 stations has dwindled to just 15. Measuring water is key to helping people prepare for natural disasters and adapt to climate change—so researchers are increasingly filling data gaps using information gathered from space. Read the full story.
—Maria Gallucci
We can still have nice things
A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or tweet ’em at me.)
+ If you love a good steak, here’s where you can track down some of the best.
+ Black Mirror creator Charlie Brooker reflects on something truly terrifying: the workplace.
+ Happy birthday to the one and only Prince of Darkness!
+ The bar of the HR Giger Museum in Switzerland looks exactly how you’d expect it would—completely mindblowing.
Every year, MIT Technology Review recognizes 35 young innovators who are doing pioneering work across a range of technical fields including biotechnology, materials science, artificial intelligence, computing, and more.
We’re now taking nominations for our 2025 list and you can submit one here. The process takes just a few minutes. Nominations will close at 11:59 PM ET on January 20, 2025. You can nominate yourself or someone you know, based anywhere in the world. The only rule is that the nominee must be under the age of 35 on October 1, 2025.
We want to hear about people who have made outstanding contributions to their fields and are making an early impact in their careers. Perhaps they’ve led an important scientific advance, founded a company that’s addressing an urgent problem, or discovered a new way to deploy an existing technology that improves people’s lives.
If you want to nominate someone, you should identify a clear advance or innovation for which they are primarily responsible. We seek to highlight innovators whose breakthroughs are broad in scope and whose influence reaches beyond their immediate scientific communities.
The 2025 class of innovators will join a long list of distinguished honorees. We featured Lisu Su, now CEO of AMD, when she was 32 years old; Andrew Ng, a computer scientist and serial entrepreneur, made the list in 2008 when he was an assistant professor at Stanford. That same year, we featured 31-year-old Jack Dorsey—two years after he launched Twitter. And Helen Greiner, co-founder of iRobot, was on the list in 1999.
Know someone who should be on our 2025 list? We’d love to hear about them. Submit your nomination today or visit our FAQ to learn more.
This story is from The Algorithm, our weekly newsletter on AI. To get it in your inbox first, sign up here.
President Biden first witnessed the capabilities of ChatGPT in 2022 during a demo from Arati Prabhakar, the director of the White House Office of Science and Technology Policy, in the oval office. That demo set a slew of events into motion and encouraged President Biden to support the US’s AI sector while managing the safety risks that will come from it.
Prabhakar was a key player in passing the president’s executive order on AI in 2023, which sets rules for tech companies to make AI safer and more transparent (though it relies on voluntary participation). Before serving in President Biden’s cabinet, she held a number of government roles, from rallying for domestic production of semiconductors to heading up DARPA, the Pentagon’s famed research department.
I had a chance to sit down with Prabhakar earlier this month. We discussed AI risks, immigration policies, the CHIPS Act, the public’s faith in science, and how it all may change under Trump.
The change of administrations comes at a chaotic time for AI. Trump’s team has not presented a clear thesis on how it will handle artificial intelligence, but plenty of people in it want to see that executive order dismantled. Trump said as much in July, endorsing the Republican platform that says the executive order “hinders AI innovation and imposes Radical Leftwing ideas on the development of this technology.” Powerful industry players, like venture capitalist Marc Andreessen, have said they support that move. However, complicating that narrative will be Elon Musk, who for years has expressed fears about doomsday AI scenarios and has been supportive of some regulations aiming to promote AI safety. No one really knows exactly what’s coming next, but Prabhakar has plenty of thoughts about what’s happened so far.
For her insights about the most important AI developments of the last administration, and what might happen in the next one, read my conversation with Arati Prabhakar.
Now read the rest of The Algorithm
Deeper Learning
These AI Minecraft characters did weirdly human stuff all on their own
The video game Minecraft is increasingly popular as a testing ground for AI models and agents. That’s a trend startup Altera recently embraced. It unleashed up to 1,000 software agents at a time, powered by large language models (LLMs), to interact with one another. Given just a nudge through text prompting, they developed a remarkable range of personality traits, preferences, and specialist roles, with no further inputs from their human creators. Remarkably, they spontaneously made friends, invented jobs, and even spread religion.
Why this matters: AI agents can execute tasks and exhibit autonomy, taking initiative in digital environments. This is another example of how the behaviors of such agents, with minimal prompting from humans, can be both impressive and downright bizarre. The people working to bring agents into the world have bold ambitions for them. Altera’s founder, Robert Yang sees the Minecraft experiments as an early step towards large-scale “AI civilizations” with agents that can coexist and work alongside us in digital spaces. “The true power of AI will be unlocked when we have truly autonomous agents that can collaborate at scale,” says Yang. Read more from Niall Firth.
Bits and Bytes
OpenAI is exploring advertising
Building and maintaining some of the world’s leading AI models doesn’t come cheap. The Financial Times has reported that OpenAI is hiring advertising talent from big tech rivals in a push to increase revenues. (Financial Times)
Landlords are using AI to raise rents, and cities are starting to push back
RealPage is a tech company that collects proprietary lease information on how much renters are paying and then uses an AI model to suggest to realtors how much to charge on apartments. Eight states and many municipalities have joined antitrust suits against the company, saying it constitutes an “unlawful information-sharing scheme” and inflates rental prices. (The Markup)
The way we measure progress in AI is terrible
Whenever new models come out, the companies that make them advertise how they perform in benchmark tests against other models. There are even leaderboards that rank them. But new research suggests these measurement methods aren’t helpful. (MIT Technology Review)
Nvidia has released a model that can create sounds and music
AI tools to make music and audio have received less attention than their counterparts that create images and video, except when the companies that make them get sued. Now, chip maker Nvidia has entered the space with a tool that creates impressive sound effects and music. (Ars Technica)
Artists say they leaked OpenAI’s Sora video model in protest
Many artists are outraged at the tech company for training its models on their work without compensating them. Now, a group of artists who were beta testers for OpenAI’s Sora model say they leaked it out of protest. (The Verge)
A new home for Threads tips and insights.
While Elon Musk’s wants to build an ‘everything app’, Meta’s attempts at the same have repeatedly fallen short.
TikTok Minis could offer alternative transactional options in-stream.
Meta’s looking to appease Trump’s concerns with its political efforts.