MIT Disavowed a Viral Paper Claiming That AI Leads to More Scientific Discoveries
The Massachusetts Institute of Technology (MIT) is distancing itself from a headline-making paper about AI's purported ability to accelerate the speed of science.
The paper in question is "Artificial Intelligence, Scientific Discovery, and Product Innovation," and was published in December as a pre-print by an MIT graduate student in economics, Aidan Toner-Rodgers. It quickly generated buzz, and outlets including The Wall Street Journal, Nature, and The Atlantic covered the paper's (alleged) findings, which purported to demonstrate how the embrace of AI at a materials science lab led to a significant increase in workforce productivity and scientific discovery, albeit, at the cost of workforce happiness.
Toner-Rodgers' work even earned praise from top MIT economists David Autor and 2024 Nobel laureate Daron Acemoglu, the latter of whom called the paper "fantastic."
But it seems that praise was premature, to put it mildly. In a press release on Friday, MIT conceded that following an internal investigation, it "has no confidence in the provenance, reliability or validity of the data and has no confidence in the veracity of the research contained in the paper." MIT didn't give a reason for its backpedaling, citing "student privacy laws and MIT policy," but it's a black eye on MIT nonetheless.
The university has also requested that the paper be removed from the ePrint archive ArXiv and requested it be withdrawn from consideration by the Quarterly Journal of Economics, where it's currently under review.
The ordeal is "more than just embarrassing," Autor told the WSJ in a new report, "it's heartbreaking."
According to the WSJ's latest story, the course reversal kicked off in January, when an unnamed computer scientist "with experience in materials science" approached Autor and Acemoglu with questions about how the AI tech centered in the study actually worked, and "how a lab he wasn't aware of had experienced gains in innovation."
When Autor and Acemoglu were unable to get to the bottom of those questions on their own, they took their concerns to MIT's higher-ups. Enter, months later: Friday's press release, in which Autor and Acemoglu, in a joint statement, said they wanted to "set the record straight."
That a paper evidently so flawed passed under so many well-educated eyes with little apparent pushback is, on the one hand, pretty shocking. Then again, as materials scientist Ben Shindel wrote in a blog post, its conclusion — that AI means more scientific productivity, but less joy — feels somewhat intuitive. And yet, according to the WSJ's reporting, it wasn't until closer inspection by someone with domain expertise, who could see through the paper's optimistic veneer, that those seemingly intuitive threads unwound.
More on AI and the workforce: AI Is Helping Job Seekers Lie, Flood the Market, and Steal Jobs
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
37 minutes ago
- Yahoo
Midjourney 推出首個 AI 影片生成模型 V1，正式進軍生成影片服務行列
雖然大家都經常玩 ChatGPT 的圖像生成功能，但說到元祖級、最強的 AI 圖像生成服務，必定是Midjourney，而他們在星期三宣布推出首款 AI 影片創作模型 V1，正式進軍生成影片服務行列。用戶只需上傳一張圖片或相片，就能自動生成一條約 4 – 5 秒長的影片。 上傳相片後，V1 可以很簡單地用自動方式生成影片，當然亦有提供一些設定讓用戶去調整，例如使用手動模式以文字描述想要加入的特定動畫效果，又或者調整鏡頭走向等。Midjourney V1 可由一張相片自動生成一條為 480p 解像度、約 5 秒長的影片，但其實用戶在生成後可延長影片四秒、最多四次，因此最長是可生成時長達 21 秒的影片。目前想體驗 V1 的話，每月USD $10 的 Basic 訂閱計劃就可以試用得到，而 USD $60 的 Pro 計劃與 USD $120 的 Mega 計劃用戶，則可在「Relax」模式下無限量地生成影片。Midjourney 表示將會在接下來的一個月內，重新評估影片模型的收費方案。 Introducing our V1 Video Model. It's fun, easy, and beautiful. Available at 10$/month, it's the first video model for *everyone* and it's available now. — Midjourney (@midjourney) June 18, 2025 Midjourney 對 AI 影片模型的期望，不僅於為電影領域影片提供補充素材（B-roll）或製作廣告。據 TechCrunch 的報導指，Midjourney 創始人大衛霍爾茲（David Holz）表示 AI 影片模型的下一步是建構出能夠「即時運行的開放世界模擬」的 AI 模型。現時 Midjourney 正處於與迪士尼與環球影業的侵權訴訟之中，會否成為新服務的絆腳石，屬未知之數。 更多內容： TechCrunch 迪士尼與環球影業聯合狀告 AI 製圖 Midjourney：「侵權與抄襲的無底洞」求償超過 5.9 億美元 古天樂 x AI！本地電影導入 AI 技術示範作，用 Google VEO 2 協助製高質影片 Google I/O 2025 | Google 的 Veo 3 AI 模型現在可以為影片生成搭配的音軌 緊貼最新科技資訊、網購優惠，追隨 Yahoo Tech 各大社交平台！ 🎉📱 Tech Facebook： 🎉📱 Tech Instagram： 🎉📱 Tech WhatsApp 社群： 🎉📱 Tech WhatsApp 頻道： 🎉📱 Tech Telegram 頻道：
Yahoo
42 minutes ago
- Yahoo
Better Cybersecurity Stock: CrowdStrike or SentinelOne?
CrowdStrike and SentinelOne have similar business models and offerings. CrowdStrike is much larger. SentinelOne trades at a steep discount to CrowdStrike. 10 stocks we like better than CrowdStrike › Artificial intelligence (AI) may have many benefits, but it's also making it easier for hackers, online criminals, and other digital malefactors to threaten businesses, and those threats are getting more potent. Keeping them at bay requires a lot of funds to be devoted to cybersecurity, making companies like CrowdStrike (NASDAQ: CRWD) and SentinelOne (NYSE: S) excellent investment opportunities. But is there an advantage to buying one over the other now? Both companies' base products are AI-powered protection platforms that analyze digital activity and learn to spot the threats among the normal activity. They deploy their software to network endpoints -- in other words, laptops, smartphones, and other devices that can access a client's internal network. By protecting these devices, companies make it harder for cyberattackers to gain access to their internal networks, where they might steal sensitive information, delete files, interfere with systems, or even lock them down with ransomware to extort payments from their victims. While endpoint protection is how both companies land clients, each bolsters its offerings with an array of other cybersecurity products that clients can use to create a protection suite tailored to their unique situations. Since these two direct competitors offer highly similar product types, it's hard to declare either a winner on this front from an investor perspective. Winner: Tie. From a sheer size perspective, CrowdStrike is the clear winner. During its fiscal 2026 first quarter, which ended April 30, CrowdStrike's annual recurring revenue (ARR) rose to $4.4 billion. SentinelOne's ARR of $948 million in its fiscal Q1 was less than a quarter of that. While size doesn't always matter, in this case, it does. Because so many more companies use CrowdStrike's platform, it's more likely that any given IT professional will have at least one contact already on its client list. If CrowdStrike is doing a great job with those clients, word will spread, and it will likely receive more serious consideration in future cybersecurity bidding processes. This advantage cannot be understated. Indeed, it's one of the reasons why CrowdStrike's growth has remained strong despite its size. Winner: CrowdStrike In terms of growth rates, SentinelOne is slightly outperforming CrowdStrike in this category. However, this should be no surprise because SentinelOne is a much smaller company. In fiscal Q1, SentinelOne's ARR rose 24% year over year, while CrowdStrike's increased 22% year over year. While I will give the point to SentinelOne, it's important to understand that CrowdStrike is growing from a much larger base than SentinelOne, making this close call all the more impressive for CrowdStrike. Winner: SentinelOne Due to its smaller size and focus on top-line growth, SentinelOne is far from profitable, while CrowdStrike has achieved intermittent profitability (although it reverted to a negative operating margin and a loss in its most recent quarter). SentinelOne is far from breaking even, but CrowdStrike was in this same position about five years ago. There's no reason not to expect SentinelOne to follow a similar path to profitability, but it will take some time. Meanwhile, CrowdStrike should eventually turn a profit again, as it has proven that it can do that. Winner: CrowdStrike CrowdStrike is leading this battle of the stocks so far, but SentinelOne is about to change the narrative with one jaw-dropping metric. CrowdStrike is the most popular cybersecurity stock in the market, and as a result, it has been bid up to expensive levels. From a price-to-sales (P/S) standpoint (the best metric to use to compare these companies since CrowdStrike flips between profitable and unprofitable, while SentinelOne is years away from profits), CrowdStrike has gotten far more expensive than SentinelOne over the past few years. CrowdStrike stock is now five times more expensive than SentinelOne, which is hard to believe, considering they compete in the same industry and are growing at nearly identical rates. This leads me to believe that CrowdStrike's stock has been overly hyped up while SentinelOne has been forgotten. While I'm OK with valuing CrowdStrike at a premium due to its market leadership position, this is far too great a premium to pay. SentinelOne is a dirt-cheap stock, and CrowdStrike is almost too expensive to consider. While I have been a long-term CrowdStrike bull, I'd be a bit cautious about buying the stock at its current lofty valuation. As a result, I think SentinelOne is the better cybersecurity investment right now. Before you buy stock in CrowdStrike, consider this: The Motley Fool Stock Advisor analyst team just identified what they believe are the for investors to buy now… and CrowdStrike wasn't one of them. The 10 stocks that made the cut could produce monster returns in the coming years. Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you'd have $664,089!* Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you'd have $881,731!* Now, it's worth noting Stock Advisor's total average return is 994% — a market-crushing outperformance compared to 172% for the S&P 500. Don't miss out on the latest top 10 list, available when you join . See the 10 stocks » *Stock Advisor returns as of June 9, 2025 Keithen Drury has positions in CrowdStrike. The Motley Fool has positions in and recommends CrowdStrike. The Motley Fool has a disclosure policy. Better Cybersecurity Stock: CrowdStrike or SentinelOne? was originally published by The Motley Fool Sign in to access your portfolio
Yahoo
2 hours ago
- Yahoo
Google AI is worse at Pokemon than I was when I was 5 – taking 800 hours to beat the Elite 4 and having a breakdown when its HP got low
When you buy through links on our articles, Future and its syndication partners may earn a commission. If you're someone who thinks AI is almost ready to take over the world, I have some good or bad (depending on your stance on things) news for you: Google's Gemini 2.5 Pro took over 800 hours to beat the 29-year-old children's game Pokemon Blue. There's a Twitch account called Gemini_Plays_Pokemon, a pale imitation of the incredible Twitch Plays Pokemon account that started this trend. First things first: how long did it take the AI to actually complete the game? Well, it was a staggering 813 hours. I feel like you could hit buttons randomly and beat the game faster than that. After some tweaks by the creator of this Twitch channel, the AI managed to halve its time to a still outrageous 406.5 hours. That is actually dead on half the time, which is interesting mathematically but still far too long to beat a game you can win with an overleveled Venusaur. Additionally, as spotted by our friends at PC Gamer, Google DeepMind reported on the Twitch account, and something unusual happens whenever its Pokemon get low on health or power points (PP). Whenever one or both of these conditions are met, "model performance appears to correlate with a qualitatively observable degradation in the model's reasoning capability – for instance, completely forgetting to use the pathfinder tool in stretches of gameplay while this condition persists." This, combined with the AI mistakenly thinking it was playing FireRed and LeafGreen and would need to find the Tea to progress, are part of the reasons it took so long to finish. Honestly, AI just isn't very good at playing Pokemon. Someone else made Claude Plays Pokemon, and that AI spent hours trying to get out of Cerulean city because it kept jumping down a ledge to talk to an NPC it had already spoken to dozens of times. So, these AIs aren't able to beat a game that we could when we barely knew our times tables. Let's not worry about them taking our jobs any time soon. In the meantime, check out the best Pokemon games of all time.
