AI Hype Explodes Again:
AI related companies have accrued at least $200 billion in debt and the figure is likely considerably higher because that estimate doesn’t count undisclosed private deals. Source
AI related companies have accrued at least $200 billion in debt and the figure is likely considerably higher because that estimate doesn’t count undisclosed private deals.
New AI models usually create excitement among the true believers and this year hasn’t been an exception. OpenAI and Anthropic released GPT-5.3 Codex and Claude Opus 4.6 respectively on the same day, February 5. The Opus 4.6 included new add-ons that enable Claude Code to perform a range of functions typically filled by software providers, which came on top of predictions by Anthropic’s CEO, Dario Amodei, that 50% of entry-level white-collar jobs could vanish in 5 years as AI takes over workplaces.
The market reacted very quickly. Shares of software-as-a-service companies like Adobe, Intuit, and Salesforce declined sharply on fears that AI tools might chip away at their business. Tech giants with large AI businesses like Microsoft, Amazon, and Google were also hit hard. A trillion dollars in market cap has been recently wiped out.
The Hype A recent post on X packaged these and other events into a great story, “Something Big Is Happening,” and it has received 70 million views in the last few days. The post places these events within the context of how much generative AI has changed since OpenAI released ChatGPT in 2022, and yes, those changes are remarkable.
But he leaves a lot out, unfortunately. First, generative neural networks have been around since the late 2000s, and big companies were pushing them on selected customers several years before Sam Altman released ChatGPT to everyone at prices far below costs, which is why OpenAI, Anthropic, and almost every other AI software company is losing unprecedented amounts of money.
Second, the blog cites a study by METR, “that tracks the length of real-world tasks (measured by how long they take a human expert) that a model can complete successfully end-to-end without human help. About a year ago, the answer was roughly ten minutes. Then it was an hour. Then several hours. The most recent measurement (Claude Opus 4.5, from November) showed the AI completing tasks that take a human expert nearly five hours. And that number is doubling approximately every seven months, with recent data suggesting it may be accelerating to as fast as every four months.”
Sounds, great doesn’t it? But the blog, and many citations of METR’s work, doesn’t mention that the model only has to complete the tasks with 50% accuracy, which is unacceptable for almost every job in the world.
