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. Employees at the world's leading AI labs are saying there's a real possibility that advanced AI could destroy humanity. Are they right? Or is this more scaremongering and hype?
Will Heaven
London, Malmesbury, Richmond
As seen in:
Daily Mail,
The Telegraph,
Financial Times,
Digg,
Flipboard,
Heise Online,
MIT Technology Review,
ProQuest,
The Code Project,
EIN News
and
Comms for @Dyson. Previous: Managing Ed at @Spectator, @policy_exchange Comms Director, @Telegraph journalist, Speechwriter for Michael Gove @MoJGovUK.
Is this you? As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.
Claim your profile
Articles by Will Heaven
The Download: biotech’s future and cheaper, cleaner steel
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. Every year, MIT Technology Review puts together our 35 Innovators Under 35, a list of some of the brightest and best young minds working across science and technology. This year’s honorees include nine people transforming biotech, whose work spans everything from lifesaving innovations to groundbreaking longevity tech.
Google DeepMind wants to know if chatbots are just virtue signaling
In short, the appearance of moral behavior in LLMs should not be taken at face value. Models must be probed to see how robust that moral behavior really is. “For people to trust the answers, you need to know how you got there,” says Haas. What Haas, Isaac, and their colleagues at Google DeepMind propose is a new line of research to develop more rigorous techniques for evaluating moral competence in LLMs. This would include tests designed to push models to change their responses to moral questions.
LLMs contain a LOT of parameters. But what’s a parameter?
A large language model’s parameters are often said to be the dials and levers that control how it behaves. Think of a planet-size pinball machine that sends its balls pinging from one end to the other via billions of paddles and bumpers set just so. Tweak those settings and the balls will behave in a different way. OpenAI’s GPT-3, released in 2020, had 175 billion parameters.
The new biologists treating LLMs like an alien autopsy
How large is a large language model? Think about it this way. In the center of San Francisco there’s a hill called Twin Peaks from which you can view nearly the entire city. Picture all of it—every block and intersection, every neighborhood and park, as far as you can see—covered in sheets of paper. Now picture that paper filled with numbers.
The great AI hype correction of 2025
Here are four ways to think about the state of AI at the end of 2025: The start of a much-needed hype correction. • LLMs are not everything: In some ways, it is the hype around large language models, not AI as a whole, that needs correcting. It has become obvious that LLMs are not the doorway to artificial general intelligence, or AGI, a hypothetical technology that some insist will one day be able to do any (cognitive) task a human can.
Gestehe, KI: OpenAI bringt Modell dazu, seinen Betrug zuzugeben
OpenAI-Forscher können große Sprachmodelle (Large Language Models, LLM) nach eigenen Angaben dazu bringen, ein "Geständnis" abzulegen. Dabei erklärt das LLM, wie es eine Aufgabe ausgeführt hat, und gibt – in den meisten Fällen – sein Fehlverhalten zu.
The State of AI: life in 2030
Welcome back to The State of AI, a new collaboration between the Financial Times and MIT Technology Review.
OpenAI has trained its LLM to confess to bad behavior
Chains of thought are like scratch pads that models use to break down tasks, make notes, and plan their next actions. Analyzing them can give clear clues about what an LLM is doing. But they are not always easy to understand. And as models get larger and more efficient, some researchers think that chains of thought may become terser and even harder for humans to read. Confessions are a way to get a sense of what an LLM is doing without having to rely on chains of thought.
What’s next for AlphaFold: A conversation with a Google DeepMind Nobel laureate
Think of it this way, he says.
OpenAI’s new LLM exposes the secrets of how AI really works
The experimental model won't compete with the biggest and best, but it could tell us why they behave in weird ways—and how trustworthy they really are. ChatGPT maker OpenAI has built an experimental large language model that is far easier to understand than typical models. That’s a big deal, because today’s LLMs are black boxes: Nobody fully understands how they do what they do.
OpenAI’s new LLM exposes the secrets of how AI really works
“Neural networks are big and complicated and tangled up and very difficult to understand,” says Dan Mossing, who leads the mechanistic interpretability team at OpenAI. “We’ve sort of said: ‘Okay, what if we tried to make that not the case?’” Instead of building a model using a dense network, OpenAI started with a type of neural network known as a weight-sparse transformer, in which each neuron is connected to only a few other neurons.
Why recycling isn’t enough to address the plastic problem
So, in March 2022, the UN Environment Assembly set out to develop an international treaty to address plastic pollution. Pretty much everyone should agree that a bunch of plastic waste floating in the ocean is a bad thing. But as we’ve learned over the past few years, as these talks developed, opinions diverge on what to do about it and how any interventions should happen. One phrase that’s become quite contentious is the “full life cycle” of plastic.
The US could really use an affordable electric truck
As Andrew Hawkins pointed out in The Verge this week, “Ford seems to realize its timing is unfortunate.” During the announcement, executives emphasized that this was a bet, one that might not work out. CEO Jim Farley put it bluntly: “The automotive industry has a graveyard littered with affordable vehicles that were launched in our country with all good intentions, and they fizzled out with idle plants, laid-off workers, and red ink.” Woof.
We did the math on AI’s energy footprint. Here’s the story you haven’t heard.
The emissions from individual AI text, image, and video queries seem small—until you add up what the industry isn’t tracking and consider where it’s heading next. AI’s integration into our lives is the most significant shift in online life in more than a decade. Hundreds of millions of people now regularly turn to chatbots for help with homework, research, coding, or to create images and videos. But what’s powering all of that?
Roundtables: What DeepSeek’s Breakout Success Means for AI
Speakers: Charlotte Jee, news editor, Will Douglas Heaven, senior AI editor, and Caiwei Chen, China reporter. The tech world is abuzz over a new open-source reasoning AI model developed by DeepSeek, a Chinese startup. Its success is remarkable given the constraints that Chinese AI companies face due to US export controls on cutting-edge chips. DeepSeek’s approach represents a radical change in how AI gets built, and could shift the tech world’s center of gravity.
The startup trying to turn the web into a database
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, where queries return specific and precise results. Exa already provides its search engine as a back-end service to companies that want to build their own applications on top of it.
This manga publisher is using Anthropic’s AI to translate Japanese comics into English
A Japanese publishing startup is using Anthropic’s flagship large language model Claude to help translate manga into English, allowing the company to churn out a new title for a Western audience in just a few days rather than the two to three months it would take a team of humans. Orange was founded by Shoko Ugaki, a manga superfan who (according to VP of product Rei Kuroda) has some 10,000 titles in his house. The company now wants more people outside Japan to have access to them.
How Wayve’s driverless cars will meet one of their biggest challenges yet
The UK driverless-car startup Wayve is headed west. The firm’s cars learned to drive on the streets of London. But Wayve has announced that it will begin testing its tech in and around San Francisco as well. And that brings a new challenge: Its AI will need to switch from driving on the left to driving on the right. As visitors to or from the UK will know, making that switch is harder than it sounds.
AI trained on AI garbage spits out AI garbage
“You can certainly imagine that the same happens with machine learning models,” he says. “So if the first model has seen half of the internet, then perhaps the second model is not going to ask for half of the internet, but actually scrape the latest 100,000 tweets, and fit the model on top of it.” Additionally, the internet doesn’t hold an unlimited amount of data. To feed their appetite for more, future AI models may need to train on synthetic data—or data that has been produced by AI.
Why does AI hallucinate?
The World Health Organization’s new chatbot launched on April 2 with the best of intentions. A fresh-faced virtual avatar backed by GPT-3.5, SARAH (Smart AI Resource Assistant for Health) dispenses health tips in eight different languages, 24/7, about how to eat well, quit smoking, de-stress, and more, for millions around the world. But like all chatbots, SARAH can flub its answers. It was quickly found to give out incorrect information.
Why does AI hallucinate?
A fresh-faced virtual avatar backed by GPT-3.5, SARAH (Smart AI Resource Assistant for Health) dispenses health tips in eight different languages, 24/7, about how to eat well, quit smoking, de-stress, and more, for millions around the world. But like all chatbots, SARAH can flub its answers. It was quickly found to give out incorrect information. In one case, it came up with a list of fake names and addresses for nonexistent clinics in San Francisco.
撮れなかった写真がある-- 生成AIでよみがえる記憶
Generative AI can turn your most precious memories into photos that never existed 生成AIを用いた「シンセティック・メモリーズ(合成記憶)」プロジェクトは、世界中の家族がカメラに一度も収まることのなかった過去を取り戻す支援をしている。研究者と共同で、認知症の治療での応用も計画中だ。 by Will Douglas Heaven2024.04.11 0 マリアは1940年代にスペインのバルセロナで育った。父親との最初の思い出は鮮明だ。6歳だったマリアは、父に会いたくなると同じ建物の隣人のアパートを訪ねた。そこから、バルコニーの柵越しに下の刑務所を覗き、独房の小さな窓から、フランシスコ・フランコの独裁政権に反対して収監されていた父の姿を垣間見ようとした。 そのバルコニーに立つマリアの写真はない。しかし、マリアは今、そんな写真を手にしている。バルセロナを拠点とするデザイン・スタジオ、ドメスティック・データ・ストリーマーズ(Domestic Data...
Generative AI can turn your most precious memories into photos that never existed
Maria grew up in Barcelona, Spain, in the 1940s. Her first memories of her father are vivid. As a six-year-old, Maria would visit a neighbor’s apartment in her building when she wanted to see him. From there, she could peer through the railings of a balcony into the prison below and try to catch a glimpse of him through the small window of his cell, where he was locked up for opposing the dictatorship of Francisco Franco. There is no photo of Maria on that balcony.
A conversation with OpenAI’s first artist in residence
Alex Reben’s work is often absurd, sometimes surreal: a mash-up of giant ears imagined by DALL-E and sculpted by hand out of marble; critical burns generated by ChatGPT that thumb the nose at all AI art. But its message is relevant to everyone. Reben is interested in the roles humans play in a world filled with machines, and how those roles are changing. “I kind of use humor and absurdity to deal with a lot of these issues,” says Reben.
How environmental DNA is giving scientists a new way to understand our world
The problem is rather fundamental given eDNA’s reputation as an indirect way of detecting the undetectable — or as a workaround in contexts when it’s simply not possible to dip a net and catch all the organisms in the sea. “It is very hard to validate some of these scenarios,” Hajibabaei said.
OpenAI teases an amazing new generative video model called Sora
By signing up for our newsletter, you agree to our Terms of Use and Privacy Policy
Google’s Gemini is now in everything. Here’s how you can try it out.
By signing up for our newsletter, you agree to our Terms of Use and Privacy Policy
Bitcoin, Tech, Bio News and Insights
MIT Technology Review’s What’s Next series looks across industries, trends, and technologies to give you a first look at the future. You can read the rest of our series here. This time last year we did something reckless. In an industry where nothing stands still, we had a go at predicting the future. How did we do?
Now we know what OpenAI’s superalignment team has been up to
OpenAI has announced the first results from its superalignment team, the firm’s in-house initiative dedicated to preventing a superintelligence—a hypothetical future computer that can outsmart humans—from going rogue. Unlike many of the company’s announcements, this heralds no big breakthrough.
Google DeepMind used a large language model to solve an unsolvable math problem
By signing up for our newsletter, you agree to our Terms of Use and Privacy Policy
How AI assistants are already changing the way code gets made
By signing up for our newsletter, you agree to our Terms of Use and Privacy Policy
How AI assistants are already changing the way code gets made
For Iansiti, it’s not just about getting existing developers to produce more code. He argues that these tools will increase the demand for programmers because companies could get more code for less money. At the same time, there will be more coders because these tools lower the barrier to entry. “We’re going to see an expansion in who can contribute to software development,” he says.
A plan to redesign the internet could make apps that no one controls
Dfinity is building what it calls the internet computer, a decentralized technology spread across a network of independent data centers that allow the software to run anywhere on the internet rather than in server farms that are increasingly controlled by large firms, such as Amazon Web Services or Google Cloud. The company will be releasing its software to third-party developers, who it hopes will start making the internet computer’s killer apps. Rewinding the internet is not about nostalgia.
Wie KI-Bild- und Textgeneratoren die Kreativ-Branche umkrempeln
Ende 2021 spielte ein kleines Team von Forschern in den Büros von OpenAI in San Francisco mit einer Idee herum. Sie hatten eine neue Version eines Text-Bild-Modells entwickelt. DALL-E ist eine KI, die kurze schriftliche Beschreibungen in Bilder umwandelt: einen von Van Gogh gemalten Fuchs vielleicht oder einen Corgi aus Pizza. Jetzt mussten sie nur noch herausfinden, was sie damit anfangen könnten. Als sie an dem Modell tüftelten, wurde allen Beteiligten schnell klar, dass dies etwas Besonderes war.
解説:空前のブーム「チャットGPT」はどこから生まれたのか?
サンフランシスコに拠点を置くオープンAI(OpenAI)が12月に公開したチャットボット「チャットGPT(ChatGPT)」は、ほぼ一夜にして爆発的に話題を呼び主役となった。チャットGPTは、サービス開始からわずか2カ月後の2023年1月に、ユーザー数が1億人に到達した。史上最も急速に成長しているインターネット・サービスだという推定もある。マイクロソフトはオープンAIに100億ドルを投資し、マイクロソフト・オフィスと検索エンジンの「ビング(Bing)」にこの技術を組み込んでいる。検索を巡る戦いで再び表舞台に出てきたかつてのライバルに刺激され、グーグルは独自のAI言語モデル「ラムダ(LaMDA)」を活用したチャットボット「バード(Bard)」のリリースを急ぐ。私の家族のワッツアップ(WhatsApp)でさえ、チャットGPTに関する話で埋め尽くされているほどの盛り上がりだ。...
グーグル一強に終止符か? チャットGPTが開いた 「検索の未来」めぐる戦い
The ChatGPT-fueled battle for search is bigger than Microsoft or Google グーグルの独壇場だった検索市場に新たな時代が訪れようとしている。チャットGPTの登場を機に、巨大テック企業や新興企業が相次いで新しい検索サービスを発表しているのだ。 by Will Douglas Heaven2023.02.20 0 検索を生業とするスタートアップ企業にとって、絶好の時代がやってきた。2月上旬、筆者がユー・ドット・コム(You.com)の最高経営責任者(CEO)であるリチャード・ソッハーと話したとき、ソッハーCEOは興奮して声高に言った。「まったく、なんてエキサイティングな時代なんだろう。また新たな記録を達成したよ。こんなにも多いユーザー数は初めてだ。本当に目まぐるしい毎日だよ」。その様子からは、世界最大級の2つの企業がユー・ドット・コムのライバル製品を発表したばかりだとは思いもよらないだろう。...
Microsoft gegen Google gegen alle: Wie ChatGPT einen neuen Suchkrieg auslöste
Such-Start-ups erleben gerade eine gute Zeit. Als die US-Ausgabe MIT Technology Review kürzlich mit Richard Socher, dem CEO von You.com, sprach, war der ganz aus dem Häuschen: "Was für ein aufregender Tag – sieht nach einem neuen Rekord für uns aus", rief er. "Wir hatten noch nie so viele Nutzer. Es war wie ein Wirbelwind." Man hätte dabei nicht ahnen können, dass zwei der größten IT-Firmen der Welt gerade konkurrierende Versionen des Produkts seiner kleiner Unternehmung angekündigt hatten.
The ChatGPT-fueled battle for search is bigger than Microsoft or Google
Hoover says that Andi avoids merely repeating textual content from search outcomes. “It doesn’t make issues up like different chatbots,” she says. Folks can resolve for themselves whether or not or not that’s true. After gathering suggestions from its customers for the previous yr, the corporate’s chatbot will now typically admit when it’s not assured about a solution. “It’ll say, ‘I’m unsure, however in line with Wikipedia …,’” says Hoover.
The ChatGPT-fueled battle for search is bigger than Microsoft or Google
22 hours ago Looking to cash in on the generative AI craze, Yext, the platform for online brand management, today announced an AI-powered chatbot called Yext Chat. Taking inspiration from OpenAI’s ChatGPT, Yext Chat is designed for enterprise use cases — and differentiated, Yext claims, by a partly proprietary …
AI is dreaming up drugs that no one has ever seen. Now we’ve got to see if they work.
Immediately, on common, it takes greater than 10 years and billions of {dollars} to develop a brand new drug. The imaginative and prescient is to make use of AI to make drug discovery sooner and cheaper. By predicting how potential medicine would possibly behave within the physique and discarding dead-end compounds earlier than they go away the pc, machine-learning fashions can minimize down on the necessity for painstaking lab work.
AI is dreaming up drugs that no one has ever seen. Now we’ve got to see if they work.
At 82 years old, with an aggressive form of blood cancer that six courses of chemotherapy had failed to eliminate, “Paul” appeared to be out of options. With each long and unpleasant round of treatment, his doctors had been working their way down a list of common cancer drugs, hoping to hit on something that would prove effective—and crossing them off one by one. The usual cancer killers were not doing their job.
AI is dreaming up drugs that no one has ever seen. Now we’ve got to see if they work. [MIT Tech Review]
At 82 years old, with an aggressive form of blood cancer that six courses of chemotherapy had failed to eliminate, “Paul” appeared to be out of options. With each long and unpleasant round of treatment, his doctors had been working their way down a list of common cancer drugs, hoping to hit on something that… View Article on MIT Tech Review At 82 years old, with an aggressive form of blood cancer that six courses of chemotherapy had failed to eliminate, “Paul” appeared to be out of options.
AI is dreaming up drugs that no one has ever seen. Now we’ve got to see if they work.
1 hour ago Using a CT scanner, researchers have learned what’s been kept under wraps for all these years. For decades, a small blackened mummy has been a source of mystery for priests at a Japanese temple. The creature, which looks like a monkey-mermaid hybrid, has been kept in southern Japan’s Enjuin temple, …
Konkurrenz für KI-Videogeneratoren: Neues KI-Modell erstellt längere Videos
Stable Diffusion kennen mittlerweile viele Menschen: Mit dieser KI-Technik ist es möglich, aus Texteingaben realistische Bilder herzustellen. Das Start-up Runway, eine Firma, die an der Entwicklung dieser Technik entscheidend beteiligt war, hat nun ein generatives KI-Modell entwickelt, das auch Videos herstellen kann. Mit der Software namens Gen-1 ist es möglich, einen bestehenden Film zu nehmen und in ein anderes Video eines beliebigen Stils zu verwandeln.
ChatGPT is everywhere. Here’s where it came from
Eighties–’90s: Recurrent Neural Networks ChatGPT is a model of GPT-3, a big language mannequin additionally developed by OpenAI. Language fashions are a kind of neural community that has been skilled on tons and many textual content.
ChatGPT is everywhere. Here’s where it came from [MIT Tech Review]
We’ve reached peak ChatGPT. Released in December as a web app by the San Francisco–based firm OpenAI, the chatbot exploded into the mainstream almost overnight. According to some estimates, it is the fastest-growing internet service ever, reaching 100 million users in January, just two months after launch. Through OpenAI’s $10 billion deal with Microsoft, the… View Article on MIT Tech Review We’ve reached peak ChatGPT.
生成AIが新章突入、Stable Diffusion共同開発元が動画版を発表
生成AI(ジェネレーティブAI)のスタートアップ企業であるランウェイ(Runway)は、テキスト入力や参照画像で指定した任意のスタイルを適用して、既存の映像を新しい映像に変換できるAIモデル「Gen-1」を発表した。同社は、2022年に話題となったテキストから画像を生成するモデル「ステーブル・ディフュージョン(Stable Diffusion)」を、ステイビリティ.AI(Stability.AI)と共同開発した企業である。 ランウェイのWebサイトに掲載されているデモリールでは、Gen-1が道行く人々をクレイアニメの人形に、テーブルに積み上げられた本を夜の街並みに変える様子が紹介されている。ステーブル・ディフュージョンが画像を生成したように、ランウェイはGen-1で映像を作り出したいと考えている。ランウェイの最高経営責任者(CEO)で共同創業者のクリストバル・ヴァレンスエラは、「画像生成モデルは爆発的に普及しています。私は、2023年は映像の年になると確信しています」と語る。...
The original startup behind Stable Diffusion has launched a generative AI for video
Arrange in 2018, Runway has been creating AI-powered video-editing software program for a number of years. Its instruments are utilized by TikTokers and YouTubers in addition to mainstream film and TV studios. The makers of The Late Present with Stephen Colbert used Runway software program to edit the present’s graphics; the visible results workforce behind the hit film All the pieces All over the place All at As soon as used the corporate’s tech to assist create sure scenes.
The original startup behind Stable Diffusion has launched a generative AI for video
Our new Privacy Policy will go into effect on January 1st, 2023. Follow the link to read our updated Privacy Policy
Show More
loading
Actions
Is this you?
As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.Get in touch with Will
Contact Will, search articles and posts on X, monitor coverage, and track replies from one place.
Learn more about Muck Rack