- AI For All
- Posts
- Has the World Gone Mad?
In partnership with
Hello readers,
Welcome to the AI For All newsletter! Big Tech remains devoted to AI while investors are concerned, GPT-5 won’t ship this year, Microsoft Copilot is seeing tepid adoption, and a tsunami of AI overhype flooded the coasts of our integrity. Let’s dive in!
Has the World Gone Mad?
Prompt: Confused man staring into void
The definition of insanity is doing the same thing over and over and expecting different results. Microsoft, Meta, Alphabet, and Amazon had their Q3 earnings calls, and the story was the same for all of them: AI remains unprofitable while enormously increasing CapEx, yet all four companies plan to keep spending anyway because of AI’s supposed long-term potential. Investors were not reassured, and shares fell.
Let’s focus on MMA (Microsoft, Meta, Alphabet), Zuck’s favorite sport. Microsoft said Azure growth is likely to slow and blamed capacity constraints at its data centers. Not helping things is OpenAI, which will lose Microsoft $1.5 billion in the current period. During a Reddit AMA, Sam Altman said that the company faces “limitations and hard decisions” when it comes to allocating compute resources. Also, GPT-5 won’t ship this year, and Sora, DALL·E 4, camera mode for ChatGPT, and vision for Advanced Voice Mode are all either delayed or without release dates.
And then there’s this: [On Microsoft Copilot] "When asked whether it’s been worth the cost, tech leaders say they aren’t sure. An equal one-quarter of respondents answered either yes or no to that question, while 50% of respondents said it is still too soon to know.” This means that of the people who aren’t indecisive, 50% say Microsoft Copilot is not worth it. Marc Benioff concurs (while peddling the exact same kind of product).
Meanwhile, Sundar Pichai claimed that “more than a quarter of all new code at Google is generated by AI.” While that figure is probably exaggerated and misleading, it does help to explain the decline of Google’s entire user interface. Google Search has only gotten worse as a product, and now AI Overviews, a feature nobody asked for, is expanding to more than a 100 countries (greatly increasing the cost for Google).
Pichai says, “the feedback we’ve received for AI Overviews has been highly positive,” but he provides no evidence for this. He also says, “people prefer using Search with AI Overviews, and they find their search results more helpful.” Let me see if I understand this correctly. Google intentionally made Search worse in such a way as to increase ad revenue and is now losing billions of dollars to make it “helpful” again by adding AI instead of just making Search good again like before.
Finally, a tone-deaf Mark Zuckerberg vowed to add even more AI-generated slop to Facebook with entire AI-generated feeds. Why? To assault you with more ads, of course. Such wanton disregard is not limited to users. Advertisers have to deal with Meta’s AI-powered ad platform, which is apparently riddled with money-losing bugs.
Next, we have a few cases of shameless overhype. Microsoft wrote on their blog: “AI has reset our expectations […] From transforming how we explore the cosmos to enabling doctors to provide personalized care and making business functions operate more intelligently, it all comes down to you—the developer—to turn this potential into reality.” Translation: It’s on you to make our product useful.
Also, we hear a lot about AI’s benefits in healthcare, but researchers found that OpenAI’s Whisper model, used by many medical centers to transcribe patients’ consultations with doctors, frequently makes stuff up. Furthermore, The New England Journal of Medicine found that LLM-generated text risks degrading the medical record.
Now, these next two cases are legitimately insane. Speaking at a conference, a weary-looking Elon Musk said that AI capabilities are increasing at a rate of 100x a year and that in four years, AI will be 10,000x better than it is today. This is not even remotely true. This is a total lie. If you’ve been following this newsletter, you should be privy to the fact that LLMs are showing every sign of plateauing (among a litany of other issues).
Elon also thinks that AI “will be able to do anything any human can do possibly within the next year or two.” I have never been more certain of something being wrong in my entire life. He went on to predict that there will be 10 billion Tesla Bots that can do anything by 2040. Here’s what the world leader in robotics, Boston Dynamics, has mustered (muskered? sorry) thus far: a slow, awkward robot with a superfluous rotating torso that performs a basic task much less efficiently than a human would.
Musk, in his boorishness, sounds like a parody of AI hype as compared with his fellow travelers — like SoftBank’s Masayoshi Son, who just sounds like a confused uncle. Son claimed that artificial superintelligence will be 10,000x smarter than a human and will exist by 2035. Um, what? How were those numbers derived exactly? He also predicted that generative AI will ultimately cost $900 trillion in cumulative capital expenditure. Putting aside the insanity of that, why would this be the case? Doesn’t Son realize that scaling LLMs is not going to lead to superintelligence? That real superintelligence would be energy efficient and not require more data than exists to barely reach mediocrity?
Why do we listen to these people? We need to stop assuming that successful people must be “clever” and know something we don’t. This pathological benefit-of-the-doubt-giving by people in parasocial relationships with the rich and powerful has gotten so bad that it seems to have even been extended to dictators by certain sects of society. After all, who’s more successful than a dictator?
Lastly, the White House issued a national security memorandum on AI, and it’s filled with the same old misconceptions no doubt implanted by credulous advisors:
“Predicting technological change with certainty is impossible, but the foundational drivers that have underpinned recent AI progress show little sign of abating.” For the millionth time, not true. There’s growing reason to doubt the scaling hypothesis.
“To retain its lead in AI, the United States must continue developing the world’s most sophisticated AI semiconductors and constructing its most advanced AI-dedicated computational infrastructure.” Hardware isn’t what’s holding AI back.
“Although the United States has benefited from a head start in AI, competitors are working hard to catch up, have identified AI as a top strategic priority, and may soon devote resources to research and development that United States AI developers cannot match.” Is this starting to sound like a wild goose chase to anyone else?
🔥 Rapid Fire
OpenAI introduces ChatGPT search to challenge Google and Perplexity
OpenAI plans in-house chip with Broadcom and TSMC, abandons foundry
GitHub announces new Copilot model options and more at GitHub Universe
Apple releases Apple Intelligence on all devices to underwhelming reviews
Researchers find writing backwards can trick AI into providing bomb recipe
Google adds Grounding with Google Search to Gemini API and AI Studio
Google brings Gemini to Maps to drive you off a cliff when it hallucinates
Microsoft delays Recall again over security it should have had from the start
KKR and Energy Capital Partners invest $50 billion in AI data center projects
Receive Honest News Today
Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.
📖 What We’re Reading
“A survey of 500 Asset Performance Management (APM) practitioners, conducted by leading technology research and advisory firm ARC Advisory Group, reflects how more manufacturers recognize the business benefits of technologies like predictive maintenance (PdM) and IIoT – and how these benefits will only amplify with AI enhancements. However, the report also comes with a warning: there needs to be more focus on increasing employee satisfaction.”