Software developer employment is projected to grow by 267,700 from 2024 to 2034.
Luis Alvarez/Getty Images
Several manager, tech, and healthcare jobs pay well and could see strong job growth over the next decade.
Business Insider used new projection data to see how high-paying jobs could grow.
Software developers and financial managers typically make six figures and are projected to be in high demand.
Software developers aren't going away, the latest job projections show. This job typically pays well and will still be needed in the era of artificial intelligence.
The Bureau of Labor Statistics published new employment projections on Thursday that showed potential growth from 2024 to 2034. Business Insider took these new findings to analyze which high-paying jobs are expected to grow a lot.
To do so, we took the geometric mean of occupations that had a median annual wage of at least $75,000 in 2024 and a projected employment increase over the next decade and ranked the jobs by that combined metric. We excluded catch-all job groups.
Below are the results, along with their projected job growth from 2024 to 2034, median annual wage in 2024, and typical education needed.
20. Sales managers
Milan Markovic/Getty Images
Projected increase: 29,000
Median pay: $138,060
Typical education needed: Bachelor's degree
19. Marketing managers
Luis Alvarez/Getty Images
Projected increase: 26,700
Median pay: $161,030
Typical education needed: Bachelor's degree
18. Physician assistants
SDI Productions/Getty Images
Projected increase: 33,200
Median pay: $133,260
Typical education needed: Master's degree
17. Computer systems analysts
Erik Isakson/Getty Images
Projected increase: 45,500
Median pay: $103,790
Typical education needed: Bachelor's degree
16. Market research analysts and marketing specialists
Igor Suka/Getty Images
Projected increase: 63,000
Median pay: $76,950
Typical education needed: Bachelor's degree
15. Construction managers
boonchai wedmakawand/Getty Images
Projected increase: 48,100
Median pay: $106,980
Typical education needed: Bachelor's degree
14. Postsecondary health specialties teachers
SDI Productions/Getty Images
Projected increase: 50,100
Median pay: $105,620
Typical education needed: Doctoral or professional degree
13. Lawyers
Heide Benser/Getty Images
Projected increase: 35,900
Median pay: $151,160
Typical education needed: Doctoral or professional degree
When you buy through our links, Business Insider may earn an affiliate commission. Learn more
James Gill - Danehouse/Getty Images
The season is still young, and we can show you where to watch Premier League live streams, wherever you are in the world. There's even an option for free coverage of one of the big games each weekend.
Chelsea vs. Fulham is this week's free game via 9Now in Australia, and you can follow the instructions below to tune in from anywhere in the world for free via a VPN if you're away from the country at the moment.
Liverpool, Spurs, and Arsenal share the top of the table with nine points each. Tottenham surprised us all last week with a dominant win over City. Arsenal and Liverpool will meet on Monday for week three's biggest fixture that could see one of them top the table.
US coverage of the Premier League airs through various NBCUniversal properties, including NBC, the USA Network, and Peacock. The majority of the action will be available to live stream on Peacock, where plans start at $11 a month or $110 a year, although some games (like the August 15 kickoff) will be missing.
Cord-cutters who still want to tune into matches that aren't on Peacock or those seeking a more comprehensive live TV streaming experience can check out month-to-month services like DirecTV, Sling TV, or Fubo.
DirecTV's MySports package includes USA, NBC, and around 20 other popular live sports channels, including TNT and ESPN. The pack also includes access to ESPN Plus content. Subscriptions cost $70 a month, but new customers can get a five-day free trial. Local channels like NBC occasionally have some regional variability, so it's a good idea to check the exact channel lineup available to you before subscribing.
Sling TV carries USA, NBC, and 40+ other channels in its Blue plan. Subscriptions cost $46 a month (or $51, depending on your location in the US). You can also try getting one of Sling's new season passes for $199, which spans five months and saves you about $31 when you prepay. Local channels vary by region, so you should check your ZIP code on the Sling website to ensure you can access networks like NBC.
Fubo also carries NBC and USA in both its Pro and Elite plans. The Pro plan costs $85 a month and includes access to more than 200 channels. New users can get $30 off their first month of service after a five-day free trial.
Where to watch Premier League in the UK
The majority of Premier League matches will be available through Sky Sports in the UK. Prices vary depending on which plan and contract length you sign up for, and you can learn more in our guide to the best Sky TV deals and plans. Sky Sports is also available to stream through Now. Additional matches will air on TNT Sports, which fans can live stream on Discovery Plus for £31 a month.
Where to watch Premier League in Canada
Fubo is the must-have streaming home of the Premier League in Canada. The live TV streaming service has two tiers: Sports Basic and Sports. Sports Basic will get you 200+ Premier League matches, while Sports proper has all matches. Both tiers also include access to loads of other sports. Basic costs $20 a month, but you can ultimately get a cheaper subscription if you pay quarterly or annually. The same goes for the regular Sports plan, which costs $31.50 a month on its own but is cheaper monthly if you sign up for a longer-term plan. Right now, customers can get their first month of Sports for just $27.
How to watch Premier League from anywhere
Football fans traveling abroad can keep up with their usual streaming services from anywhere with the help of a VPN, or virtual private network. VPNs allow people to change their device's virtual location. This way, their go-to websites and apps work from anywhere in the world, so they don't need to worry about missing out on their favorite programming while traveling. Plus, they're instant ways to boost online security.
Our No. 1 recommendation is NordVPN, a top-notch VPN with a 30-day money-back guarantee. It's chock-full of cybersecurity perks and has a few savings opportunities with long-term contracts, but customers can also just subscribe month-to-month. To learn more, you can read our official NordVPN review.
Note: The use of VPNs is illegal in certain countries, and using VPNs to access region-locked streaming content might constitute a breach of the terms of use for certain services. Business Insider does not endorse or condone the illegal use of VPNs.
China's tech rebound is hit by a fierce food delivery price war among Alibaba, JD.com, and Meituan.
Consumers got rock-bottom prices, but bargains are fading as corporate profits come under pressure.
Chinese regulators have summoned the trio to demand an end to the excessive competition.
China's tech rebound, sparked by AI buzz around DeepSeek earlier this year, has been interrupted by a bruising food delivery war that hammered the shares of Alibaba, JD.com, and Meituan — three of the country's biggest e-commerce players.
The intense rivalry among the three tech giants has showered Chinese consumers with dirt-cheap indulgences — bubble tea and lattes for as little as 1 Chinese yuan, or $0.14, and meals dropped at their door in under 30 minutes.
It's not just food. As growth in China's traditional e-commerce slows, companies are racing into the new fast-delivery segment.
"It can be flowers, it can be medications, it can be toiletries," Jason Yu, the managing director for Greater China at consumer insights company Kantar Worldpanel, told Business Insider.
But the golden age of cheap convenience is nearing an end, thanks to a race-to-the-bottom price war that's even drawn the government's attention.
JD.com's official push into food delivery in February triggered the latest escalation, forcing Meituan and Alibaba to respond with deeper subsidies.
The three platforms have poured big money into discounts, eroding profits and rattling investors.
On Thursday, Meituan — which controls about half of China's food delivery market — warned of losses this quarter due to "irrational competition," sending its shares plunging as much as 13% in one day. The stock is down 33% so far this year.
JD.com reported earlier this month that its net profit had halved for the quarter ending June. The company's stock is down 12.3% this year.
Alibaba is posting its fiscal first-quarter earnings later on Friday, but the rivalry has also affected its share price. The New York-listed stock is down about 20% from a March peak, though it is still up 41% this year to date.
The price war may have benefited consumers, but it has also drawn scrutiny from Beijing. Regulators are cracking down on excessive competition as the world's second-largest economy battles a yearslong property crisis and deflation.
In May and June, China's top market regulator summoned Alibaba, Meituan, and JD.com to demand an end to "disorderly competition." In response, the trio pledged restraint.
End of an era
Consumers may now be facing the end of an era, as one-yuan coffees and near-free meals give way to more realistic delivery costs.
"The low prices will not persist as the business model is fundamentally unsustainable, with companies burning billions in losses," said MingYii Lai, a strategy consultant at Daxue Consulting.
"Both financially and also politically, I don't think it's in their interest to continue that aggressive subsidy war," said Kantar Worldpanel's Yu.
Now, the most extreme promotions — like near-free burgers or tea — have started to fade, though weekend discounts remain common, Yu said.
Yu said subsidies are likely to vanish within the next year or two, as platforms count on consumer behavior becoming entrenched.
"Once the consumer's habit is there, then actually all these subsidies will be withdrawn, and then consumers will have to pay a fair price for delivery," he said.
Amazon struggles to attract AI talent due to its pay model and perception of falling behind others.
Amazon's compensation model has long caused complaints from employees.
Competitors like Meta and OpenAI offer more attractive packages for AI engineers.
As the AI talent war sweeps across Silicon Valley, Amazon has largely sat on the sidelines. A confidential internal document, and accounts from people familiar with the matter, reveal why.
The company has flagged its unique pay structure, lagging AI reputation, and rigid return-to-office rules as major hurdles. Now, the tech giant is being pushed to rethink its recruiting strategy as it scrambles to compete for top talent.
The document, from late last year, was written by the HR team covering Amazon's non-retail businesses, including Amazon Web Services, advertising, devices, entertainment, and the newly formed artificial general intelligence team.
"GenAI hiring faces challenges like location, compensation, and Amazon's perceived lag in the space," the document noted. "Competitors often provide more comprehensive and aggressive packages." Business Insider obtained a copy of the document.
Amazon's absence from recent splashy AI hires underscores those concerns. Meta has pulled in high-profile talent from ScaleAI, Apple, and OpenAI. Google and OpenAI continue to be top destinations for AI experts, while Microsoft has even drafted a wish list of Meta AI employees it hopes to recruit.
Amazon's spokesperson initially told BI that the company continues to "adapt our approach to remain highly competitive, maintaining flexibility in both our compensation packages and work arrangements to attract and retain the best AI talent in this dynamic market."
Hours later, the spokesperson updated the statement, saying the premise of the story was "wrong," without providing any specifics.
"We continue to attract and retain some of the best people in the world and they're building and deploying GenAI applications at a rapid clip. Our compensation is competitive, but we also want missionaries who are passionate about inventing things that will make a meaningful difference for customers — for those kinds of people, there's no better place in the world to build."
Door desks and 'egalitarian' pay
Amazon founder Jeff Bezos back in the 1990s
TNS/ABACA via Reuters Connect
Amazon is famously frugal. One of its origin stories recounts how the company bought cheap doors from Home Depot and hacked them together as office desks. This became guiding symbol of Amazon's cautious spending, with founder Jeff Bezos still using door desks today.
This penny-pinching culture has smashed straight into an AI hiring battle that's being fueled by unprecedented spending, putting Amazon in a tricky situation.
The internal document described compensation as one of the "hotly debated topics" among Amazon recruiters, citing the company's strict use of fixed salary bands for each role. Amazon's "egalitarian philosophy" on pay leaves its offers "below par" compared with top rivals, it added.
"The lack of salary range increases for several key job families over the past few years does not position Amazon as an employer of choice for top tech talent," the document warned.
For Amazon, missing out on top AI talent is a potential risk. The pool of top-tier AI researchers and engineers is limited, and without experts with deep knowhow, it's hard to compete at the frontier of the field. Indeed, Amazon has yet to find a blockbuster AI product like OpenAI's ChatGPT or Anthropic's Claude, although its Bedrock AI cloud service has made progress.
Amazon's pay structure has been a long-standing source of tension.
Several people who spoke to Business Insider cited the 2020 departure of Amazon robotics VP Brad Porter as evidence of the company's frugal approach hampering talent recruitment and retention. Porter left in part after Amazon refused to raise his pay band.
Amazon's stock vesting schedule is also heavily backloaded, a structure that can be less attractive to new hires. The policy extends even to top executives, who generally receive no cash bonuses.
'Voting with their feet'
Amazon CEO Andy Jassy
REUTERS/Brendan McDermid
In addition to highlighting Amazon's "perceived lag in the AI space," the internal document said generative AI has further intensified the competition for specialized talent, particularly individuals with expertise in large language models.
An August report from venture capital firm SignalFire shows Amazon is on the lower end of engineering retention, far below Meta, OpenAI, and Anthropic. Jarod Reyes, SignalFire's head of developer community, told Business Insider that Amazon rivals are making bigger strides in AI, across open models, foundational research, and developer tooling.
"Amazon hasn't clearly positioned itself as a leader in the generative AI wave," Reyes said. "Engineers are paying attention and they're voting with their feet."
SignalFire chart on engineering talent retention
SignalFire
Some investors share that view. On Amazon's earnings call last month, Morgan Stanley analyst Brian Nowak pressed CEO Andy Jassy on Wall Street's "narrative right now that AWS is falling behind" in AI and fears of losing market share to rivals. Jassy's response fell flat, sending Amazon's stock lower during the call.
Amazon intends to tackle these concerns. According to the document, the company will refine its "compensation and location strategy" and host more events designed to highlight its generative AI capabilities. It also intends to set up dedicated recruiting teams for generative AI within business units like AWS to boost efficiency.
'Hubs' constrain talent
Hundreds of tech workers gather outside the Amazon Spheres during a Climate Strike walkout and march in Seattle, Washington, U.S. September 20, 2019.
REUTERS/Lindsey Wasson
Another point of contention is Amazon's aggressive return-to-office mandate, which has already caused logistical issues.
The company's new "hub" policy — which requires employees to relocate to a central office or risk termination — has further limited its access to "high-demand talent like those with GenAI skills," according to the internal document.
"Hubs constrain market availability," it stated.
Amazon is exploring ways to allow for more "location-flexible" roles, the document added.
Amazon hasn't been entirely on the sidelines. Last year, it brought on Adept CEO David Luan as part of a licensing deal with the AI startup. Luan now heads Amazon's AI agents lab. But the company has also seen departures, including senior AI leaders including chip designer Rami Sinno and VP Vasi Philomin, who worked on Bedrock.
One Amazon recruiter told Business Insider that a growing number of job candidates started declining offers last year because of the company's RTO policy. Even if a competitor pays less, people are open to taking the job if they can stay remote, this person said.
"We are losing out on talent," this person added.
Indeed, Bloomberg reported recently that Oracle has hired away more than 600 Amazon employees in the past two years because Amazon's strict RTO policy has made poaching easier.
Staying the course
The internal Amazon document dates to late last year, leaving open the possibility that the company has since adjusted its compensation approach to make exceptions for top AI talent.
Still, multiple people familiar with the situation told Business Insider there haven't been any formal updates to internal pay guidelines. One current Amazon manager said it remains almost impossible for the company to enact sweeping changes, given its long track record of sticking to the existing system. The people who spoke with Business Insider asked not to be identified discussing sensitive matters.
"Based on how we run our business and what we have achieved, there are more risks than potential benefits from changing an approach that has been so successful for our shareholders over the past several decades," Amazon wrote this year about executive compensation in its annual proxy statement.
Of course, the AI talent war may end up being an expensive and misguided strategy, stoked by hype and investor over-exuberance.
Some of the high-profile recruits Meta recently lured have already departed.
Urban Outfitters' CEO said the company will not suffer too much from the closure of the de minimis loophole.
Associated Press
Urban Outfitters isn't too concerned about the closing of the de minimis loophole.
The company's chiefs said the new rule would lower the sales of competitors like Shein.
President Donald Trump is set to close the tax loophole for goods from all countries on Friday.
Urban Outfitters said President Donald Trump's closing of the de minimis loophole will not hurt the company.
The retail giant's chief executives said in an earnings call on Wednesday that the de minimis loophole, which until now allowed small parcels under $800 to enter the US duty-free, would hit competitors like Shein.
Francis Conforti, Urban Outfitters' operating chief, said the exemption had "really immaterial impact" on the company and that he was not worried that its absence would impact the business.
CEO Richard Hayne added that the closure of the loophole would "only help" Urban Outfitters.
"Some of the folks who were big into this, Shein, and some others, are obviously having a little bit harder time coping with some of the new regulations," Hayne said to investors. "So to the degree that they're shipping less, it should help us."
Urban Outfitters, a Philadelphia-based corporation, owns several retailers under its brand umbrella, such as Anthropologie, Free People, and Nuuly.
The company reported total sales of $1.5 billion in the latest quarter, an 11.3% increase compared to last year. Its stock price is up about 42% since the start of the year.
Shein, a budget fast-fashion retailer from China, was one of the first to be affected by Trump's closing of the de minimis exemption for goods coming from China in May. The US is Shein's biggest market.
In April, it announced that it would be increasing its prices, as its operating expenses had gone up because of the changes in trade rules.
According to an October report by US Customs and Border Protection, de minimis shipments accounted for 92% of all cargo entering the US. The report added that at that point, the agency processed about 4 million de minimis shipments a day, up from 2.8 million in 2023.
Urban Outfitters, while unharmed from the de minimis closure, has not escaped Trump's trade changes unscathed. In an earnings call in May, the company said it would be bringing fall inventory in sooner, anticipating supply chain issues from Trump's tariffs.
Representatives for Urban Outfitters and Shein did not respond to requests for comment from Business Insider.
Artificial Societies has secured $5.35 million in seed funding led by Point72 Ventures.
The startup creates AI personas for product testing and brand propositions.
We got an exclusive look at the pitch deck it used to secure the fresh funding.
London-based Artificial Societies, which was part of Y Combinator's Winter 2025 batch, has secured $5.35 million in seed funding.
The startup, which launched in 2024, has developed a platform that generates AI personas that help companies test how their product, marketing, branding, and other content might be perceived.
"We primarily use AI agents that chat with each other, mimicking how real humans influence each other," cofounder and CEO James He told Business Insider. "We try to capture that artificial society, and how it will respond to any input. It can be a marketing message or a product proposition, and we produce analysis for our customers."
The startup sells its market research services to individual and enterprise customers. Looking ahead, He said there's scope for the startup to apply the technology in policymaking.
"In the long term, there's huge potential to use this to evaluate different policies before we spend more effort on drawing it out, and on public consultation," he said.
The fundraising environment was tricky to navigate as first-time founders, He added, but the team used their own technology to simulate the investment process. He told Business Insider this meant they were "quite prepared" when it came to positioning their startup while pitching to VCs.
The seed round was led by Point72 Ventures, with additional participation from angels associated with Google DeepMind, Strava, and Sequoia Scout.
With the fresh funding, Artificial Societies plans to expand its research talent for its simulation engine and double down on its distribution efforts.
We got an exclusive look at the 10-slide deck used to secure the funding.
Samantha Gold started her children's clothing brand, Motette, after moving to Sydney.
Courtesy of Samantha Gold
Samantha Gold started a children's pajama brand called Motette when her family moved to Sydney.
She typically starts work at 3 a.m. because of the time difference from the East Coast.
The brand is growing rapidly, but she says her work hours are unsustainable.
This as-told-to essay is based on a conversation with Samantha Gold, founder and CEO of children's pajama brandMotette. It has been edited for length and clarity.
In 2024, my husband got a job opportunity in Sydney. We were living in Charleston, SC, at the time, and I was the CEO ofElizabeth James The Label.
Starting my own business had been on my mind for a while, and it felt like a great time to do it when we moved. I quit my job, and my husband and I decided to use the proceeds from selling our house to fund my new clothing company. I worked for months figuring out what I'd need to make it work and then officially launched my children's pajama brand, Motette, in March 2025. At the time, I thoughtstarting a businessfocused on selling to US customers while I was living in Australia would be challenging but doable.
After a year of working from roughly 3 a.m. to 4 p.m. each day, though, I've realized it's unsustainable. The business is doing well, but we're moving our familyback to the Statesin the coming months so I can have more of a normal work life while I continue to grow the brand.
Motette's pajamas are made from bamboo, and the designs are inspired by Gold's love of interior design.
Courtesy of Samantha Gold
I couldn't find pajamas for my kids that I liked
My kids are 2 and 3, and I never felt like I could find pajamas for them that matched my own style. I didn't want slogans, bold colors, or sequins on their pajamas, which is where the idea for the company came from.
My love ofinterior designdrives the prints on the pajamas. People often tell me that our prints look like they could be beautiful wallpaper. All of the pajamas are made from bamboo, which is super comfortable, UV-resistant, and sustainable.
Outsourcing work is crucial
Because mybusiness is self-funded, I've been really scrappy about building it. I don't have any full-time employees right now. Instead, I work with eight women from the Philippines, who help me with some of the day-to-day work on my business.
One doessocial media managementand community engagement for me, which is a core part of my strategy. She messages women on Instagram each day to tell them about the brand, and it's worked really well. I also outsource things like email marketing, wholesale outreach to boutiques we'd like to be in, and social ads.
Working odd hours has its pros and cons
Time zones can be an issue, as Australia and the East Coast don't overlap much. Right now, 4 p.m. on the East Coast is 6 a.m. in Sydney, so Istart work around 3 a.m.my time most days. That's the only way that I can overlap more than a couple of hours with potential and existing customers and wholesalers.
When I start work, I typically have a full inbox to work through. In some ways, this makes my day easier. I get through all my calls and emails in the first portion of my day and then have time for the creative side of the business after the East Coast is offline.
But it also means that by 4 p.m. my time, I'm usually exhausted. I don't have a lot of energy left when my kids get home from day care, and I'm not my best self around them as a result.
Though the business is growing rapidly, the hours Gold works are not sustainable.
Courtesy of Samantha Gold
The business is growing, but my schedule is unsustainable
I'm proud of what the brand has accomplished in just over four months. I'm a go-getter and have made myunorthodox schedulework for the past year. We've been growing steadily, and the feedback from customers has been great. We just landed contracts with Maisonette and Bloomingdale's Online, too.
If I didn't have kids or a family, this type of graveyard shift might work. But I haven't been able to be as present with my kids as I want to be, and my health has started to suffer. My skin has been breaking out horribly, and I'm stressed all the time. I'vestopped drinking coffeeand am seeing a naturopath to help me regulate my nervous system, but so far, it isn't working.
My husband has been great throughout all of this, but he's had to carry an uneven balance with our kids the past year as I've gotten the business off the ground. He's had to get them off to day care each morning, and when they wake up in the middle of the night, he always has to be the one to get them since I'm working.
My husband agrees it's the best move for our family. He's in tech sales, and we both believe there will bejob opportunitiesfor him in Austin, TX, where we're moving.
Even though I love living in Sydney, I know I'll be happier and be able to be a better version of myself and a better parent when my work schedule is more normal.
The world's leading tech companies are in a race to AGI. There is growing evidence that large language models, which underpin the most popular chatbots, might never get there.
J Studios/Getty Images
Top AI companies are in a race to develop artificial general intelligence.
The large language models powering popular chatbots, however, are showing their limits.
Some researchers say world models or other strategies might be the clearer path to AGI.
The race to build artificial general intelligence is colliding with a harsh reality: Large language models might be maxed out.
For years, the world's top AI tech talent has spent billions of dollars developing LLMs, which underpin the most widely used chatbots.
The ultimate goal of many of the companies behind these AI models, however, is to develop AGI, a still theoretical version of AI that reasons like humans. And there's growing concern that LLMs may be nearing their plateau, far from a technology capable of evolving into AGI.
AI thinkers who have long held this belief were once written off as cynical. But since the release of OpenAI's GPT-5, which, despite improvements, didn't live up to OpenAI's own hype, the doomers are lining up to say, "I told you so."
Principal among them is perhaps Gary Marcus, an AI leader and best-selling author. Since GPT-5's release, he's taken his criticism to new heights.
"Nobody with intellectual integrity should still believe that pure scaling will get us to AGI," he wrote in a blog post earlier this month, referring to the costly strategy of amassing data and data centers to reach general intelligence. "Even some of the tech bros are waking up to the reality that 'AGI in 2027' was marketing, not reality."
Here's why some think LLMs are not all they are cracked up to be, and the alternatives some AI researchers believe are the better path to AGI.
The AI bubble
OpenAI is now the most valuable startup on the planet. It has raised about $60 billion, and a discussed secondary share sale could push the company's valuation over $500 billion. That would make OpenAI the most valuable private company in the world.
There are good reasons for the excitement. According to the company, ChatGPT has 700 million weekly users, and OpenAI's products have largely set the pace of the AI race.
There are a couple of problems, however. First, and perhaps foremost for its investors, OpenAI is not profitable and shows few signs of becoming profitable soon. Second, the company's founding mission is to develop AGI in a way that benefits all of humanity, yet there's a growing feeling that this world-changing technology, which props up much of the hype around AI, is much further away than many engineers and investors originally thought.
Other companies, too, have been riding this hype wave. Google, Meta, xAI, and Anthropic are all attracting and pouring billions of dollars into scaling their LLMs, which means snapping up talent, buying data, and building vast arrays of data centers.
The mismatch between spending and revenue, and hype and reality, is provoking alarm that the AI industry is a bubble on the verge of bursting. OpenAI CEO Sam Altman himself thinks so.
"Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes. Is AI the most important thing to happen in a very long time? My opinion is also yes," he told journalists earlier this month.
While other tech leaders, like former Google CEO Eric Schmidt, are less certain, a $1 trillion stock market tech sell-off last week showed the concerns are widespread. The market recovered on Friday after Federal Reserve Chair Jerome Powell said he is considering a rate cut in September.
Now, everyone is eagerly anticipating Wednesday's earnings report from Nvidia, which makes the chips powering LLMs and is the pick-and-shovel company of the AI rush. If the company's earnings show signs of slowing and its outlook is more cautious, there will be a whole new round of worry, and the AI doomers will again remind everyone of what they've been saying for years: LLMs are not the way.
The problem with LLMs
In June, Apple researchers released a paper called "The Illusion of Thinking." What they found sounded positively human: Advanced reasoning models give up when faced with more complex tasks.
Their conclusion, however, was that these models rely on pattern recognition rather than logical thinking, and the researchers cautioned against the belief that they could result in AGI. "Claims that scaling current architectures will naturally yield general intelligence appear premature," the researchers wrote.
The paper was widely mocked online, largely because Apple, despite its size and vast resources, is perceived as far behind in the AI race. For skeptics, however, it was validating.
Andrew Gelman, a professor of statistics and political science at Columbia University, has argued that the level of textual comprehension shown by LLMs falls short of expectations. What LLMs do compared to what humans do is the difference between "jogging and running," Gelman wrote in a 2023 blog post.
"I can jog and jog and jog, thinking about all sorts of things and not feeling like I'm expending much effort, my legs pretty much move up and down of their own accord … but then if I need to run, that takes concentration," he wrote.
Geoffrey Hinton, the Nobel Prize winner known to some as the Godfather of AI, disagrees. "By training something to be really good at predicting the next word, you're actually forcing it to understand," he told The New Yorker back in 2013.
Another potential problem with LLMs is their tendency to misinterpret the meanings of words, hallucinate, and spread misinformation. This reality is why, for now, most companies adopting AI require a human in the mix.
In a report published earlier this year, a group of academic researchers in Germany specializing in computational linguistics surveyed "in-the-wild" hallucination rates for 11 LLMs across 30 languages. They found that the average hallucination rate across all languages varied between 7% to 12%.
Leading AI companies like OpenAI have, in recent years, operated under the belief that these problems can be mitigated by feeding LLMs more information. The so-called scaling laws, which OpenAI researchers outlined in a 2020 paper, state that "model performance depends most strongly on scale."
However, recently, researchers have begun to question whether LLMs have hit a wall and are facing diminishing returns as they scale. Yann LeCun, Meta's chief AI scientist who heads a lab under the company's superintelligence unit, is largely focused on next-generation AI approaches instead of LLMs.
"Most interesting problems scale extremely badly," he said at the National University of Singapore in April. "You cannot just assume that more data and more compute means smarter AI." Apple's analysis also found that current LLM-based reasoning models are inconsistent due to "fundamental limitations in how models maintain algorithmic consistency across problem scales."
Alexandr Wang, the head of Meta's superintelligence division, appears equally uncertain. He said scaling is "the biggest question in the industry" at the Cerebral Valley conference last year.
Even if scaling worked, access to high-quality data is limited.
The hunt for unique data has been so fierce that leading AI companies are pushing boundaries — sometimes at the risk of copyright violations. Meta once considered acquiring publisher Simon & Schuster as a solution. Anthropic collected and scanned millions of pirated books while training Claude, which a district judge ruled in June did not constitute fair use.
"Language doesn't exist in nature," Fei Fei Li, the Stanford professor famous for inventing ImageNet, said on an episode of Andreessen Horowitz's podcast in June. "Humans," she said, "not only do we survive, live, and work, but we build civilization beyond language."
LeCun's gripe is similar.
"We need AI systems that can learn new tasks really quickly. 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," he said during his talk in April.
New ways to AGI
Researchers like Li and LeCun are pursuing an alternative to LLMs, called world models, that they believe is a better path to AGI.
Unlike large language models, which determine outputs based on statistical relationships between words and phrases, world models make predictions by simulating and learning from the world around them. These kinds of models feel more akin to how humans learn, while LLMs rely on vast troves of data that humans have no access to.
Computer scientist and MIT professor Jay Wright Forrester outlined the value of this kind of model all the way back in a 1971 paper.
"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," he wrote. "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."
Recent research has found that world models not only capture reality as it is, but can also simulate new environments and scenarios.
In a 2018 paper, researchers David Ha and Jurgen Schmidhuber built a simple world model inspired by humans' cognitive systems. This was used to not only model hypothetical scenarios, but also to train agents.
"Training agents in the real world is even more expensive," the authors wrote. "So world models that are trained incrementally to simulate reality may prove to be useful for transferring policies back to the real world."
In August, Google's DeepMind released Genie 3, a world model that it says "pushes the boundaries of what world models can accomplish." It can model physical properties of the real world, like volcanic terrain or a dimly lit ocean. This could allow AI to make predictions based on what it learns from these real-world simulations.
There are other ideas in the works, too. Neuroscience models try to mimic the processes of the brain. Multi-agent models operate on the theory that multiple AIs interacting with each other is a better analogy to how humans function in real life. Researchers pursuing multi-agent models believe AGI is more likely to emerge through this kind of social exchange.
Then, there is embodied AI, which adapts world models into physical forms, allowing robots to interpret and train on the world around them. "Robots take in all kinds of forms and shapes," Li said on the No Priors podcast in June.
The potential of these alternatives, and in particular world models, gives hope to even Marcus, the premier LLM doomer. He refers to world models as cognitive models and urges AI companies to pivot from LLMs and focus on these alternatives.
"In some ways, LLMs far exceed humans, but in other ways, they are still no match for an ant," Marcus said in a June blog post. "Without robust cognitive models of the world, they should never be fully trusted."
Nearly two dozen people 60 and older attended an introductory AI class in August.
Clark Hodgin for BI
Amid the clanging sounds of construction from outside, some of New York City's most ambitious AI adopters aged 60 and over made their way into a 10:15 a.m. class, equipped with notepads and the occasional walker.
The Tuesday Intro to Chatting with AI class run by Senior Planet, an AARP subsidiary, had about 20 people. Most had at least some exposure to AI. Some came to learn how to apply AI at their jobs. Others wanted to go neck-and-neck with their grandchildren on ChatGPT.
The trainer, Brian Rapsey, began his lesson by asking the students if they could define AI. Three gave enthusiastic answers based on what they had read online or what their children told them.
"Every single industry is being affected by AI," Rapsey told the students, leading to some nods and looks of doubt.
In interviews for my series on people working into their 80s, dozens said they use AI at work or home, whether for maintaining client lists, seeking coding advice, or planning a vacation to Mexico. While a few skeptics felt it was a disease to society, many expressed enthusiasm for how they could get ahead in their careers or social circles. It's a sentiment that challenges many of the commonly held stereotypes about older Americans not embracing technology.
A 2024 survey from the employment nonprofit Generation found a stark age gap in attitudes toward AI-related hiring. While 57% of U.S. hiring managers said they would be very likely to hire 25- to 34-year-olds for jobs that frequently use AI tools, only 7% said the same for candidates 65 and older. More than half admitted they would be unlikely, or entirely unwilling, to consider older applicants for these roles. However, there are some promising signs in the workforce, said Mona Mourshed, CEO of Generation.
Some of the students opted to take notes with pen and paper.
Clark Hodgin for BI
"The great news is that there is a population of mid-career, older workers who not only are power users, but they are actually seeing the productivity benefits, and they're enjoying their work more," Mourshed said. "They are figuring out how to make the most of AI, and we should make the most of them in the workplace."
Ricardo Strobert, 72, sat in the front row. The retired music teacher said that he's used AI for some basic tasks, but often needs to ask for help. He told me he plans on using AI for checking out properties in his area and doing historical research.
The class introduced the students to ChatGPT, Copilot, and Gemini, briefly explaining each one's advantages. Zooming in on his search bar, Rapsey typed in ChatGPT, gave the students a warning not to use impersonator apps, and clicked on the link.
Rapsey spoke about AI's "amazing" capabilities, noting that though it may not be capable of something at the moment, it may be in a few months. He explained the difference between predictive and generative AI, giving the example of students using AI to write college essays or play chess. One student said that she heard of people using AI for pickup lines.
Marisa Giorgi, director of curriculum development at Older Adults Technology Services, which organizes the Senior Planet classes, said when the first class debuted in 2023, most students were curious about AI but had some concerns, particularly about identifying what's real. The courses have expanded to AI for small business owners, AI for communication, and AI detection.
"AI is ubiquitous right now. You see it and hear it everywhere. So older adults want to be part of that conversation, and they want to understand in a lot of cases how they could start adopting AI," Giorgi said. She added, "When you're working with older adults, and one has a lifetime of experience, there's a lot of knowledge being shared during our classes."
Sandra Smith, 69, who attended the class, retired last November from her role as an executive assistant at a lobbying firm for New York's hospitals. Though she considers herself "fairly tech savvy," she said the more she knows, the better.
"I want to stay updated on tech because I don't watch television anymore. Everything I do is online, from paying rent to hanging out," Smith said. "I found Gemini and ChatGPT, and it's like a whole other world has opened up."
The students were shocked by some of AI's capabilities.
Clark Hodgin for BI
Testing out AI functions
Rapsey asked the students to prompt ChatGPT. One woman asked it to create a funny thank-you note for her friends in their 60s who took her out for her birthday. The result drew a few laughs, as ChatGPT made jokes about "proving once again that age is just a number" and wrote, "Here's to more nights where we can stay up way past our usual bedtimes of 9:30 p.m." Rapsey encouraged the students to keep prompting it, such as asking it to add names of the friends, or making sure to include as much information as possible in the first prompt.
One woman asked how many responses AI can give to a prompt if multiple people ask it the same thing. Rapsey explained that AI will always create a unique response. Another woman asked if AI picks up on writing style, which Rapsey said it doesn't unless prompted to. Still, he said, people use different AI tools that make the writing sound as though it were written by a human.
The students were particularly fascinated by how AI could respond to photos, such as asking ChatGPT to visually identify a problem with a printer. Others were drawn to how AI could be used to help with writing books, composing music, or illustrating art.
David Pace, 72, hopes to use AI for writing and art projects. He suspects AI may help him be employable if the economy worsens or he decides to move.
Pace has known since graduate school in the early 1980s that he always needed to stay on top of technological advances. He recalled a professor telling him that if he didn't know anything about computers, he wouldn't survive. During his career as a teaching and guidance counselor, he tried to improve his computer and phone skills. When he saw the AI class, he signed up in hopes of starting his AI journey and then taking more advanced classes and studying on his own.
In a demonstration, Rapsey verbally asked ChatGPT about the top things to do in New York City, and after a few seconds, a woman's voice told of concerts and parks. He explained that the AI had to search the internet for the answer, which took a few seconds, and that it linked to its sources. Rapsey then cautioned that because AI will always answer a question, sometimes the answer will be odd or not particularly helpful.
Some students brought up AI concerns they had from reading articles, such as people losing their data after ChatGPT updated to version 5, or whether Grok was reliable. One asked if there is a dark web equivalent of AI. Another asked how well AI answers opinion-based questions. Misinformation doesn't just go away because it's AI, Rapsey warned.
Applying AI to real life
Some students wanted to use AI to help get back into the workplace.
Clark Hodgin for BI
After showcasing how to use AI, Rapsey asked the class how they might use it in their lives. One woman said she asked it to plan a five-day vacation that doesn't require driving, then prompted it again to give locations that she wouldn't know about. Another planned to use it for recipes, such as how to make a dish without a certain ingredient.
The students were perhaps most enthused by a demonstration of how far AI has come over the last few years. Rapsey showed them an AI video of Will Smith eating from two years ago and compared it to a similar one in 2025, and the students picked up on how the more recent video seemed very close to real.
The introductory class didn't go too deep into any particular topic, though it was still enlightening for many students. Some said after the class that they wanted to use AI at work, while others said AI may motivate them to return to work after retiring.
Eileen Rose, 72, said she hardly knew anything about AI before the class but felt compelled to join because she got laid off from her job earlier this year and has been on the job hunt. She considered retirement after working for nearly two decades in the financial administration department of a biotech startup, but she suspects she'll want to return to work if a company bites.
She tried AI for the first time two days before the class to write a birthday card for a friend who is into gardening. She said the class opened her eyes to how she could better optimize her resume using ChatGPT, and she plans to take more advanced classes. She finds herself "old school" and prefers in-person demonstrations.
"I figured this is dramatically changing the world, so let me at least find out about it because I hopefully have 30 more years left," Rose said. "A lot has changed in the job market, and I feel almost like a Luddite."
Pop Mart announced mini Labubu plushies and keychains.
Pop Mart
Labubu is set to get smaller and cuter.
Pop Mart announced a new line of miniature The Monsters keychains.
Small enough to hang on mobile phones, the new charms will be released on August 29.
The small plush doll with the creepy grin that has had the world in a chokehold for the past year is set to get even smaller.
On Friday, Chinese toymaker Pop Mart announced the launch of mini Labubu dolls in its new "The Monsters Pin For Love Series."
According to a release from Pop Mart, the little Labubus will cost $22.99. The dolls will be around 4 inches in height and small enough to hook onto mobile phones. For reference, an iPhone 16 has a height of 5.8 inches.
People looking to land the little Labubus should mark their calendars for August 29. The dolls will come in 30 different colors.
These mini dolls, like their larger counterparts, will also be sold in blind boxes — which means collectors will only know which color they pulled after opening their boxes.
The mini Labubu dolls each have a letter stitched on their backs.
Pop Mart
"The Monsters Pin For Love" series also includes a set of 30 letter charm pendants, each with a unique pattern and a metal Labubu charm. The letter charms are about 4.5 inches in height, per the product listing on Pop Mart's website.
This charm series will also be sold in blind boxes, and will be priced at $18.99, per Popmart's press release.
Pop Mart is launching letter pendants with Labubu metal charms.
Pop Mart
The "Pin for Love" series "allows collectors to spell names, initials, or secret messages," the release said. Fans can purchase the mini Labubus and letter charms from Pop Mart's website starting August 29 and choose shipping or in-store pickup.
Pop Mart's CEO, Wang Ning, teased the launch of mini Labubus during an earnings call this week.
The company released its first-half 2025 earnings on Tuesday. It reported a 204% increase in revenue in the first half of the year compared to the year before, with global sales of 13.87 billion Chinese yuan, or about $1.94 billion. It also reported a 401% increase in profits compared to the year before.
The Monsters IP, which includes fan favorite Labubu, contributed 4.81 billion Chinese yuan to the company's total sales in the first half of the year, per its earnings report. Pop Mart said its results were unaudited.
The company, listed in Hong Kong, has seen its stock price rise about 18% in the last five days and more than 550% in the past year.
Labubus have spiked in popularity over the past year. Desperate to get their hands on the dolls, which are released via unannounced drops, fans around the world have formed snaking queues outside Pop Mart stores.
To curb queues, the company has had to halt physical Labubu sales in some countries, such as the UK and South Korea. The doll has become so popular that its bootleg cousin, Lafufu, enjoyed its own cultural moment.
