Loading summary
Transcript1 lines
- [00:04]
Brian McCullough
Welcome to the TechMe Ride Home for Thursday, June 19th, 2025. I'm Brian McCullough. Today, are we about to see a summer of layoffs in Silicon Valley? MidJourney's new video model Meta continues its aqua hire spree with some folks we know Microsoft has its own nuclear option with OpenAI. While OpenAI is starting to get worried about its models being nuclear dangerous and what even is AGI, here's what you missed today in the world of tech if the AI job apocalypse is coming, might we see it in Silicon Valley first? Quoting Bloomberg Microsoft is planning to axe thousands of jobs, particularly in sales, as part of the company's latest move to trim its workforce amid heavy spending on artificial intelligence. The cuts are expected to be announced early next month, following the end of Microsoft's fiscal year, according to people familiar with the matter. The reductions won't exclusively affect sales teams and the timing could still change, said the people, who requested anonymity to discuss a private matter. The terminations would follow a previous round of layoffs in May that hit 6,000 people and fell hardest on product and engineering positions, largely sparing customer facing roles like sales and marketing. In April, the company told employees it planned to use third party firms to handle more sales of software to small and mid sized customers. Microsoft has said it regularly re evaluates the organizational structure to make sure it's investing for growth as the company spends tens of billions of dollars on servers and data centers. Executives have pledged to Wall street and warned employees that it would keep a lid on spending in other areas, end quote. So that doesn't sound like it's AI taking those jobs specifically, although might it be the arrival of agents doing the work of sales folks. Still, my speculative point stands. If AI is about wringing out efficiency, which again read as money, I bet Silicon Valley behemoths will be the first to midjourney has launched V1, its first AI video generation model, letting subscribers animate images via its website. Video jobs will cost eight times more than image jobs, apparently, but if you check my socials Ryan McC, you can see some of the experiments I did with this this morning. Quoting VentureBeat starting today, Midjourney's nearly 20 million users can animate images via the website, transforming their generated or uploaded stills into five second long clips, with options for extending the generation longer up to 20 seconds in five second bursts and guiding them with text. Midjourney's new offering extends its familiar image based workflow, including its new V1 text to image model. Users generate a still image either within the midjourney platform or by uploading an external file. Then press Animate to turn that image into a video. Two primary modes exist. One uses automated motion synthesis, while the other lets users write a custom motion prompt to dictate via text how elements should move in the scene so midjourney videos arrive with support for both image to video and text to video. Edits and modifications From a creative standpoint, users can toggle between two motion settings. There's a low motion mode that's optimized for ambient or minimalist movement, such as a character blinking or a light breeze shifting scenery, and high motion mode that attempts more dynamic animation of both subject and camera, though this can increase the chance of visual errors. While this is relatively short form, the company has indicated that video duration and features will expand in future updates. Meanwhile, in the AI race, the information is reporting that Meta is in advance Talks to hire friends of the pod, Nat Friedman and Daniel Gross, to work on its AI efforts and possibly buy out part of their venture capital fund. If the talks are successful, Gross would leave Safe Superintelligence, which he co founded with former OpenAI chief scientist Ilya Suskever last year. At Meta, Gross is expected to work mostly on AI products, while Friedman's remit is expected to be broader. Both Gross and Friedman are expected to work closely with Meta CEO Mark Zuckerberg and Scale AI CEO Alexander Wang, whose hiring by Meta was finalized last week in a $14.3 billion deal. As part of the talks, Meta is discussing buying out a substantial portion of NFDG's holdings and cashing out the fund's limited partners in the process, the person familiar with the discussions said. The social media giant will have minority stakes in the startups that NFDG has invested in, which could include ssi, but it will not get information about and control over the startups, the person said. Although financial terms of discussions couldn't be learned, a partial buyout of the fund could cost more than $1 billion. Friedman has been involved with Meta's AI efforts for at least the past year. In May 2024, he joined an advisory group to consult with Meta's leaders about the company's AI technology and products. He earlier ran GitHub from 2018 to 2021 after it was acquired by Microsoft. Earlier this year, Mark Zuckerberg asked Friedman to lead Meta's AI efforts altogether, the person familiar with the discussion said. Friedman declined, but helped brainstorm other candidates, including Wang. While Zuckerberg was skeptical Wang would leave Scale, Friedman convinced him a deal was possible, said a second person with knowledge of the discussions. As the Wang hiring came together, Zuckerberg approached Friedman again. This time, Friedman agreed to a deal of his own. He is currently expected to report to Wang, who is roughly 20 years his junior. Both men will be part of a small group of Meta leaders that Zuckerberg refers to as his management team, or M team. For Gross, the talks with Meta put him in an awkward position with ssi, a startup formed with the goal of building a leading AI company insulated from short term commercial pressures. The startup hasn't yet launched a product or described in detail what it planned to build. Gross's departure for Meta would damage an important investment for some top venture capital firms. In April, SSI raised $2 billion at a $32 billion valuation from investors such as Green Oaks, Andreessen Horowitz and Lightspeed Venture Partners. It has also raised money from Sequoia Capital. Together, Friedman and have invested in some of the buzziest AI startups, including search startup Perplexity and robotics startup the bot Company. The firm had more than $2 billion of assets under management as of last year, though that figure is likely higher now with the increase in value of some of their startups. End quote. Meanwhile, sources tell the FT that Microsoft is prepared to walk away from its high stakes talks with OpenAI if they cannot agree on critical issues such as the size of Microsoft Microsoft's stake in this eventuality. Microsoft would rely on its existing commercial contract to retain access to OpenAI's technology until 2030 unless there was an offer that was equal to or better than its current arrangements, according to those people. These people stressed, however, that Microsoft was operating in good faith and both parties were meeting daily to try to put a plan on the table and were hopeful a deal could be reached. In discussions over the past year, the two sides have battled over how much equity in the restructured group Microsoft should receive in exchange for the more than $13 billion it has invested in. Open discussions over the stake have ranged from 20% to 49%. The pair are also revising the terms of its wider contract, first drafted when Microsoft invested $1 billion into OpenAI in 2019. Under its current arrangement, Microsoft has exclusive rights to sell access to OpenAI's models and receives a 20% share of revenues up to $92 billion. Microsoft is reluctant to give ground on its continued access to OpenAI's technology or its share of the group's revenues, according to multiple people close to the discussions. The Wall Street Journal reported this week that considered a nuclear option of accusing Microsoft of anti competitive behavior over its partnership. Holding out is Microsoft's nuclear option and they are just making OpenAI sweat, said one person close to OpenAI, who also argued access to the ChatGPT maker's IP was necessary for Microsoft to maintain its position in the race to commercialize AI against rivals such as Google and Meta. One person close to Microsoft said the status quo was acceptable for the big tech company and that it was happy with the current contract and prepared to run it through until 2030. The market cares about how much revenue Microsoft is making, not about how much equity it owns on OpenAI, and this deal moves revenue away from Microsoft, said another person who has discussed the negotiations with Microsoft executives. The question is, what does Microsoft get in return for giving up the right to that revenue? OpenAI's chief executive Sam Altman and its chief financial officer Sarah Fryer, have also said the company is struggling to access the computing power needed to run ChatGPT, which has raced to 500 million weekly active users worldwide while also training new models and launching products. Two former Microsoft executives involved in managing OpenAI's compute requirements said the relationship between the groups had frayed significantly over the issue, particularly around Altman's demands for faster access to even more infrastructure. Even if the issues are resolved, the transaction will have to be approved by attorneys general in Delaware and California. The for profit conversion is also subject to a legal challenge from ex AI chief Elon Musk, which has been supported by former OpenAI employees. End Quote There's a new way to have a great time with friends without the booze or the bad decisions. Cornbread Hemp's THC Seltzers just came out and you've got to try them. I've got them in my fridge right now. Finally, a THC Seltzer that tastes and feels amazing. Perfect for summertime fun time. This is a low calorie drink with only 30 calories and 5 grams of sugar. Made with pure THC and all natural ingredients, no synthetics. Each can has 5 milligrams of THC, which is the perfect amount so you don't feel couch locked or paranoid. Perfect for unwinding, kicking back and enjoying the moment without alcohol or a hangover. Four delicious flavors to choose from. Blueberry Breeze, Peach Iced Tea, Raspberry Limeade and Salted Watermelon. Every single flavor I've tried is legit delicious right now. Techmeme Ride Home listeners can save 30% off their first order and enjoy free shipping on orders over 75 bucks. Head to cornbreadhemp.com ride and use code ride at checkout. That's cornbreadhemp.com ride code ride cornbread hemp this is the good life. If you bypass security to save yourself some time, you're probably not alone. But fortunately, with 1Password Extended Access Management, security and productivity no longer have to be at odds. Work happens everywhere. We send emails from our phones and edit documents on our personal tablets and laptops. 1Password Extended Access Management ensures that whatever device your employees use for work, it's healthy and uncompromised. For example, did you know most ransomware attacks stem from unmanaged devices? That's why 1Password's device trust solution blocks unsecured and unknown devices before they access your company's apps. And don't worry, 1Password still protects against the biggest attack source Compromised Credentials. Its industry leading password manager helps employees create strong, unique logins for Every app. With 1Password you get access to detailed reports on vulnerable passwords, employee device health and credentials affected by data breaches so you can extend protection beyond the office walls. Check compliance off your list with the ability to enforce policy and identity safeguards so only trusted users on secure devices can gain access to business Data. I've trusted 1Password for all my businesses for over a decade now. They're amazing. Join me go to 1Password.com ride to secure every app, device and identity, even the unmanaged ones. Right now, my listeners get a free two week trial at 1Password.com Ride all lowercase. That's 1Password.com Ride meanwhile meanwhile, and this is a pretty big meanwhile OpenAI has warned that its upcoming models could pose a higher risk of enabling the creation of biological weapons, and says it is stepping up testing of such models, quoting Axios. OpenAI executives told Axios the company expects forthcoming models will reach a high level of risk under the company's preparedness framework. As a result, the company said in a blog post it is stepping up the testing of such models as well as including fresh precautions designed to keep them from aiding in the creation of biological weapons. OpenAI didn't put an exact time frame on when the first model to hit that threshold will launch, but head of Safety Systems Johannes Heidecke told Axios, we are expecting some of the successors of our O3 reasoning model to hit that level, end quote. OpenAI isn't necessarily saying that its platform will be capable of creating new types of bioweapons, rather that it believes that without mitigations, models will soon be capable of what it calls novice uplift or a allowing those without a background in biology to do things potentially dangerous. We're not yet in the world where there's like novel, completely unknown creation of bio threats that have not existed before, heidecke said. We are more worried about replicating things that experts already are very familiar with, end quote. When it released Claude 4 last month, anthropic said it was activating fresh precautions due to the potential risk of that model aiding in the spread of biological and nuclear threats. Various companies have also been warning that it's time to start preparing for a world in which AI models are capable of meeting or exceeding human capabilities in a wide range of tasks. OpenAI is also looking to expand its work with the US national labs and the government more broadly, OpenAI policy chief Chris Lehane told Axios. We're going to explore some additional type of work that we can do in terms of how we potentially use the technology itself to be really effective at being able to combat others who may be trying to misuse it, lehane said. Lehane added that the increased capability of the most powerful models highlights the importance, at least in my view, for the AI build out around the world for the pipes to be really US led, end quote. Finally today, and I guess also meanwhile, the Financial Times takes a look at the general lack of consensus on what artificial general intelligence, or AGI Even, is, whether LLMs are the best path toward this goal, and what the tech might look like if or when it arrives. Meta's chief AI scientist Yann LeCun, considered one of the godfathers of the technology, also does not like the term AGI on the basis that human intelligence is really not that general. We're very specialized, and certainly computers can solve certain tasks much better than we can, he said recently. Instead, to describe machines that surpass human intelligence, he prefers to use artificial superintelligence, or asi. But just as Silicon Valley cannot agree on what precisely AGI or ASI is, there is also a lack of consensus on what the technology will look like if and when it arrives. Deep Mind coined the term to describe what it called AI that is at least as capable as a skilled adult at most cognitive tasks. But that definition invites more questions. How do you define a skilled adult? How do we know when we have reached most cognitive tasks? What are those tasks? In the first place? The intangibility of AGI has allowed it to become many things to many people. For some people it's a scientific goal. For some people it's a religion. And for some people it's a marketing term, says Francois Chollet, a former software engineer at Google. Well, a key feature, according to Chen, is generality, the ability to perform tasks across a wide variety of domains. They should be fairly autonomous and shouldn't need a lot of hand holding to get tasks done, he adds. AGI will have the power to bring the things that are in your head into existence very quickly, chen says, explaining that it has the potential to help people create not just images or text, but entire apps. But critics argue this definition falls short of describing a truly intelligent system. That's just automation, which is something that we've done for decades, says Chollet, the former Google engineer. DeepMind takes a different approach. I believe that typical human performance is the most natural, practical and useful way to define the minimum requirements for an AI to be considered an AGI, says DeepMind's Leg, who popularized the term artificial general intelligence early in the 2000s, when it was seen as a fringe belief. A big problem with many definitions of AGI is that they do not specify clearly enough what an AI system needs to be able to do in order to qualify as an AGI. DeepMind's definition being as capable as a skilled adult at most cognitive tasks makes it fairly obvious. Obvious, Leg says. Ask yourself if some task is a cognitive thing that people can typically do. If it is, then the AI must be able to do it to be an AGI. But then there's the question of if the current models are how we get there. OpenAI and Anthropic believe scaling up language models with more data and compute is the best path. OpenAI's new O3 model, for instance, excels at step by step reasoning across tasks like coding and math, prompting some, like economist Tyler Cohen, to claim, I know it when I see it. Others, like the researcher Mark Chen, suggest true AGI will require autonomous agents capable of taking reliable actions, innovating and even coordinating, like human organizations. Self improvement AI writing and upgrading its own code would also be key. But critics argue today's models only simulate reasoning and remain prone to hallucinations. As we mentioned, it was last week that Apple researchers announced they had found these models collapse on complex tasks. Alternative approaches, like Meta's, world models that learn from video and robotics aim to better mirror reality. Meanwhile, data shortages and the need for synthetic training data pose new challenges. And then there's this quote some people question whether artificial general intelligence is even something we should be striving for in the first place. We are yet to have a full grasp of intelligence in biology or psychology or even education, says Margaret Mitchell, the chief ethics scientist at open source AI startup Hugging Face who co wrote a paper arguing that AGI should not be seen as a North Star. Intelligence as a concept is ill defined and it's problematic, she adds. Shooting for it is a bit fraught because it functions to give an air of positivity, of goodness. It provides us with a cover of something good when in fact it's not actually a concrete thing, and instead provides a narrative about moving forward whatever technology we want to move forward. End quote Hey, I want to give a quick shout out to the listener who helped me with my Google AdWords issues in terms of thanking them for their help. Let me tell you about their service and I'm going to quote from what they asked me to share. This is not a paid ad for them exactly. This is me just thanking them for helping me gratis. QUOTE D3Data Driven Decisions is the UK based, award winning agency that cracked my Google Ads Conversion Tracking like absolute pros, these specialists live for solving complex marketing and analytics puzzles. Mixing tech expertise with behavioral science and serious curiosity. From performance marketing to tracking fixes and conversion rate optimization, they deliver clarity, insights and real results. Search D3 data driven decisions or find them on LinkedIn because insights should drive results, not confusion. End quote thanks again.