AI leaders have a new term for the fact that their models are not always so intelligent

As academics, independent developers, and the biggest tech companies in the world drive us closer to artificial general intelligence — a still hypothetical form of intelligence that matches human capabilities — they've hit some roadblocks. Many emerging models are prone to hallucinating, misinformation, and simple errors.
Google CEO Sundar Pichai referred to this phase of AI as AJI, or "artificial jagged intelligence," on a recent episode of Lex Fridman's podcast.
"I don't know who used it first, maybe Karpathy did," Pichai said, referring to deep learning and computer vision specialist Andrej Karpathy, who cofounded OpenAI before leaving last year.
AJI is a bit of a metaphor for the trajectory of AI development — jagged, marked at once by sparks of genius and basic mistakes.
In a 2024 X post titled "Jagged Intelligence," Karpathy described the term as a "word I came up with to describe the (strange, unintuitive) fact that state of the art LLMs can both perform extremely impressive tasks (e.g. solve complex math problems) while simultaneously struggle with some very dumb problems." He then posted examples of state of the art large language models failing to understand that 9.9 is bigger than 9.11, making "non-sensical decisions" in a game of tic-tac-toe, and struggling to count.
The issue is that unlike humans, "where a lot of knowledge and problem-solving capabilities are all highly correlated and improve linearly all together, from birth to adulthood," the jagged edges of AI are not always clear or predictable, Karpathy said.
Pichai echoed the idea.
"You see what they can do and then you can trivially find they make numerical errors or counting R's in strawberry or something, which seems to trip up most models," Pichai said. "I feel like we are in the AJI phase where dramatic progress, some things don't work well, but overall, you're seeing lots of progress."
In 2010, when Google DeepMind launched, its team would talk about a 20-year timeline for AGI, Pichai said. Google subsequently acquired DeepMind in 2014. Pichai thinks it'll take a little longer than that, but by 2030, "I would stress it doesn't matter what that definition is because you will have mind-blowing progress on many dimensions."
By then the world will also need a clear system for labeling AI-generated content to "distinguish reality," he said.
"Progress" is a vague term, but Pichai has spoken at length about the benefits we'll see from AI development. At the UN's Summit of the Future in September 2024, he outlined four specific ways that AI would advance humanity — improving access to knowledge in native languages, accelerating scientific discovery, mitigating climate disaster, and contributing to economic progress.

Try Our AI Features

Explore what Daily8 AI can do for you:

Learn with AIFact CheckingText to SpeechAI Summary

Related Articles

Android Authority

31 minutes ago

• Android Authority

Google Pixel Stand (2nd Gen) deal: We've never seen it this cheap!

Robert Triggs / Android Authority As Android fans, the Google Pixel Stand (2nd Gen) is obviously one of our favorite wireless chargers. It doesn't often go on sale, and the lowest we've seen it go for was $53.93 back in 2022! Today, it's on sale for just $43.99, making this a record-low price. Get the Google Pixel Stand 2nd Gen for just $43.99 ($35.01 off) This offer is available from eBay, but the seller is actually Best Buy. This gives us a level of confidence, as it is not just a random online seller. Google Pixel Stand Wireless Charger (2nd Gen) Google Pixel Stand Wireless Charger (2nd Gen) Much more than just fast wireless charging for the Pixel 7 and Pixel 6. The Pixel Stand (2nd gen) is built to unlock your Pixel 7's Assistant capabilities while fast charging. It also wirelessly charges a wide range of other smartphones quickly too. See price at eBay Save $35.01 The Google Pixel Stand (2nd Gen) is an excellent wireless charger, especially if you use a Pixel smartphone with it. It can charge Pixel devices at up to 23W, but those speeds don't work with all devices. For example, you can charge a Pixel 9 Pro XL at 23W, but the Pixel 9 Pro speeds are reduced to 21W, and the Pixel 9 can juice up at 15W. For non-Pixel phones, the maximum charging speed is 15W. It's not the fastest charger out there, but it is pretty speedy, especially with compatible Pixel devices. Robert Triggs / Android Authority The integrated fan makes it more efficient, as it keeps your device cool while charging. This means charging will throttle less, as phones usually slow down charging when they get too hot, and wireless charging produces a lot of heat. Additionally, batteries get slowly damaged with heat, so keeping the phone cooler will extend the battery's overall life. I also happen to love its clean, minimalist design. Not to mention the fact that it doubles as a stand, offering easy access to your apps and content even while charging. It's great for desks, side tables, and so on. This is a really nice deal if you want to catch Google's best wireless charger. Make sure to act quickly, though, as this is a record-low price on a product that seems to be going away soon. Google doesn't even have it in stock anymore, so chances are retailers are getting rid of extra units.

The Hill

41 minutes ago

• The Hill

Pope Leo's AI warning

The Big Story Pope Leo XIV sounded the alarm this week over artificial intelligence (AI)'s potential impact on young people's intellectual and neurological development, building upon one of the main focuses of his papacy. © Vatican Media via Associated Press The first American pope delivered the message to an annual conference on AI and ethics, part of which took place in the Vatican this week. 'All of us, I am sure, are concerned for children and young people, and the possible consequences of the use of AI on their intellectual and neurological development,' the pope said. 'Our youth must be helped, and not hindered, in their journey towards maturity and true responsibility.' The pontiff described how new generations have unprecedented, quick access to data and information, while noting they must not confuse this access to data with intelligence. 'In the end, authentic wisdom has more to do with recognizing the true meaning of life, than with the availability of data,' Leo said. Leo acknowledged AI has been used in positive ways, pointing to research in health care and science. But he warned of the possibility the emerging technology can be used for 'selfish gain at the expense of others' or 'to foment conflict and aggression.' The pope said AI, especially generative AI, 'also raises troubling questions of its possible repercussions on humanity's openness to truth and beauty, on our distinctive ability to grasp and process reality.' Since being elected pope in early May, Leo has made the risks of AI on humanity a key priority of his papacy. In his first official address to cardinals in May, the pope warned of the challenges AI poses to the 'defense of human dignity, justice and labor.' The mission resembles that of Pope Leo XIII, who pushed for workers' rights amid the industrial revolution, and a Vatican spokesperson told media outlets the name choice was 'not a casual reference.' Read more in a full report at Welcome to The Hill's Technology newsletter, we're Miranda Nazzaro and Julia Shapero — tracking the latest moves from Capitol Hill to Silicon Valley. Did someone forward you this newsletter? Subscribe here. Essential Reads How policy will be impacting the tech sector now and in the future: Meta, Oakley partner for AI-powered glasses Meta, the parent company of Facebook and Instagram, is partnering with Oakley to launch new artificial intelligence (AI) powered glasses, the tech giant announced Friday. The new smart glasses, like Meta's earlier Ray-Ban glasses, are equipped with a built-in camera, open-ear speakers and AI capabilities. The Oakley glasses also feature a longer-lasting battery and a higher resolution camera that can shoot 3K video. … Telegram founder planning to leave fortune to his 100+ children The founder of the instant messaging app Telegram said he plans to leave his multibillion-dollar fortune to the more than 100 children he has fathered either naturally or through sperm donations. Pavel Durov, in an interview published Thursday with French political magazine Le Point, said he does not differentiate between the six children he fathered naturally in three relationships and the dozens of others he fathered by sperm … Aflac discloses cybersecurity incident Aflac said Friday that it experienced a cybersecurity incident last week that may have impacted files containing social security numbers, health information and other personal information. The insurance company first detected suspicious activity on its network last Thursday and 'promptly initiated our cyber incident response protocols and stopped the intrusion within hours,' according to a press release. 'Importantly, … The Refresh News we've flagged from the intersection of tech and other topics: Crypto Corner Coinbase secures EU crypto license © Richard Drew, Associated Press Welcome to Crypto Corner, a daily feature focused on digital currency and its outlook in Washington. Crypto exchange Coinbase has obtained a license to operate in the European Union (EU) under the bloc's relatively new regulatory framework for digital assets. The company announced Friday that it has secured a Markets in Crypto Assets (MiCA) license, allowing it to operate in all 27 European member countries. Coinbase also plans to establish its European crypto hub in Luxembourg, after receiving its license from the country's Commission de Surveillance du Secteur Financier (CSSF). 'This is a pivotal moment for Coinbase, Luxembourg, and Europe's growing crypto ecosystem,' the exchange said in a press release. 'We look forward to continuing to contribute to Europe's vibrant economy, unlocking growth, and increasing economic freedom across the region.' As crypto firms seek MiCA licenses in Europe, the U.S. continues its slow forward march toward establishing its own regulatory framework. The Senate voted 68-30 Tuesday to pass the GENIUS Act, a bill creating regulatory rules of the road for one form of cryptocurrency, known as stablecoins. President Trump threw his weight behind the legislation Wednesday, urging the House to quickly pass the bill without any major additions. However, his call stands at odds with the hopes of some in Congress and the industry, who had been hoping to tie stablecoin legislation to another bill laying out a regulatory framework for the rest of the crypto market. In Other News Branch out with other reads on The Hill: ChatGPT use linked to cognitive decline: MIT research ChatGPT can harm an individual's critical thinking over time, a study released this month suggests. Researchers at MIT's Media Lab asked subjects to write several SAT essays and separated subjects into three groups — using OpenAI's ChatGPT, using Google's search engine and using nothing, which they called the 'brain‑only' group. Each subject's brain was monitored through electroencephalography (EEG), … What Others are Reading Two key stories on The Hill right now: Senate parliamentarian knocks pieces out of Trump's megabill Senate Parliamentarian Elizabeth MacDonough has ruled that several key pieces of the massive bill to implement President Trump's agenda run afoul of … Read more Former Clinton campaign chief on Democrats: 'We're leaderless, we're messageless, we're agendaless' Democratic strategist Patti Solis Doyle said the party lacks a leader, message and agenda when asked about the state of the Democratic Party in an … Read more You're all caught up. See you next week!

Yahoo

44 minutes ago

• Yahoo

OpenAI's Altman slams Mark Zuckerberg, ignites drama

OpenAI's Altman slams Mark Zuckerberg, ignites drama originally appeared on TheStreet. The race for Artificial General Intelligence (AGI) continues getting spicier. Except it's a lot less like a sprint, but more like a high-stakes soap opera starring the biggest minds in the tech world. 💵💰💰💵 The relentless push spans big-money bets and headline-making hires, all of which are part of a broader play to avoid coming in second. Needless to say, there's plenty on the line here. However, the public mud-slinging between AI leaders, especially over talent, is what's catching everyone off guard. It appears the gloves are off in the AI talent war, and the industry's needle-movers are swinging for the fences. The space is seemingly shifting from hype to execution, and everyone's looking to see who cracks AGI first. AI powerhouse OpenAI's ChatGPT lit the fuse three years ago, when suddenly, AI could chat, code, and riff like a 2023, we saw its models getting smarter, quicker, and more creative, with the likes of Google, Anthropic, and Meta Platforms () joining the race. That ushered in the open-source surge, proving top-tier AI wasn't just at the mercy of Big Tech. Now the focus is on AGI-superintelligent systems that can think, learn, and outsmart humans across the board. To put things in perspective, Superintelligent AI, or AGI, isn't just your regular chatbot. It's a transformative force that may revolutionize everything from drug discovery to enterprise software. For tech giants like Meta Platforms () , this may entail embedding intelligence across their ecosystem. From their ubiquitous social platforms to VR worlds, this could create new revenue streams and defend against disruption. Unsurprisingly, the stakes go way beyond just market share. Governments view AGI as a national security asset. Retaining the industry's cream, coupled with regular breakthroughs on home turf, has become geopolitical. More On Meta Platforms Veteran portfolio manager raises eyebrows with latest Meta Platforms move Google plans major AI shift after Meta's surprising $14 billion move WhatsApp is making a change that users won't like Hence, the victors won't just dominate tech in this ongoing war for AI's future; they could potentially reshape the global economy as we know it. Consequently, a talent war is heating up, and the company that cracks AGI first could dominate the decade. OpenAI CEO Sam Altman didn't hold back. In a podcast hosted by his brother, he revealed that Meta Platforms dangled $100 million signing bonuses to lure OpenAI staff, only to fail. 'None of our best people have decided to take them up on that,' he AI trailblazer framed the offers as mostly desperate and misguided. 'I don't think that's going to set up a great culture,' Altman said, taking a dig at Meta's compensation-first recruiting strategy. He also discussed OpenAI's edge in the fast-evolving race for superintelligence and long-term value. 'People look at the two paths and say…OpenAI has got a really good shot,' he added. Altman believes a great team is built on a collective purpose and the aim to accomplish something revolutionary. He also slammed Meta's AI efforts to date 'have not worked as well as they hoped,' suggesting a lack of an innovative culture might be the chief reason. The remarks came just days after Meta committed $14.3 billion to Scale AI and tapped its CEO, Alexandr Wang, to lead a new superintelligence project. The project aims to achieve AGI, transcending normal human intelligence. Moreover, reports suggest Meta CEO Mark Zuckerberg set out to recruit 50 top engineers and researchers to form his dream team. Perhaps a big part of Meta's recent AI push likely stems from its lukewarm reception compared to ChatGPT and DeepMind's models. After shelling out billions on his metaverse vision, Zuckerberg pivoted, chasing the next big thing in AI to stay in the game. If the plan works out, Meta's superintelligence play could reshape it all, covering everything from content to complex tasks to even an advanced version of the metaverse. It's a moonshot move, but one that could pay many dividends down the road. To bring that to life, it's already made some huge moves. Perhaps the biggest is its $14.3 billion bet on leading data labelling and annotation provider Scale AI for roughly 50% of the business. That's Meta's largest deal since its WhatsApp acquisition for $19 billion (the largest in the company's history). Scale AI's 28-year-old spunky CEO, Alexander Wang, will lead the new superintelligence team. Zuckerberg needed a sharp, Altman-style operator to fast-track AGI and win the commercial AI race. Also, it's important to note that Meta's talent hunt didn't stop at OpenAI. Of late, it has gone straight for DeepMind's core, snagging big names like Koray Kavukcuoglu (a key architect at Google). Meta also successfully poached Dr. Jack Rae, a principal researcher known for cutting-edge models from Google DeepMind. Also, it successfully tapped voice AI expert Johan Schalkwyk from startup Sesame Altman slams Mark Zuckerberg, ignites drama first appeared on TheStreet on Jun 18, 2025 This story was originally reported by TheStreet on Jun 18, 2025, where it first appeared. Sign in to access your portfolio