The AI Hype Machine: $80B for What Exactly? 🤔

Microsoft is burning through cash equivalent to 8 James Webb telescopes on AI next year. Meanwhile, LLMs still can't figure out what to do with your expense reports. Let's dig into the reality check..

Dear Friend,

The other day, a client of ours asked me the eternal question of “what’s going to change – and when?” The answer is, of course, lots of things are constantly changing – with the pace of change continuing to accelerate. And yet, so many things stay the same or just take a long time to transmorph from one state to the next. Take the good, old (well, not so old, really) smartphone – there are more than 7 billion smartphones in the world. It will take a while for these devices to be replaced by whatever might come next.

With that being said, here is what’s new… 😉

---

Headlines from the Future

No, AI Won't Take Your Job, but... ↗

A good reminder from Laurie Voss about the implications of AI on the job market – it’s the same argument Andrew Ng has been making for years now:

Jobs are more than collections of tasks. Jobs require prioritization, judgement of exceptional situations, the ability to communicate ad-hoc with other sources of information like colleagues or regulations, the ability to react to entirely unforseen circumstances, and a whole lot of experience. As I said, LLMs can deal with a certain amount of ambiguity and complexity, but the less the better. Giving them a whole, human-sized job is way more ambiguity and complexity than they can handle. It's asking them to turn text into more text. It's not going to work.

Source.

—//—

LLMs Are Good at Transforming Text Into Less Text ↗

Here is a good reminder from Laurie Voss on what LLMs are actually good at:

This is the biggest and most fundamental thing about LLMs, and a great rule of thumb for what's going to be an effective LLM application. Is what you're doing taking a large amount of text and asking the LLM to convert it into a smaller amount of text? Then it's probably going to be great at it. If you're asking it to convert into a roughly equal amount of text it will be so-so. If you're asking it to create more text than you gave it, forget about it.

Source.

—//—

Putting AI Investment Into Perspective ↗

You have heard about the enormity of the current investment landscape for AI many times before, but it’s still helpful to put the numbers into perspective (if for nothing else than to see how crazy this whole world is).

Jack Clark, in his newsletter Import AI, recently delved into Microsoft’s latest announcement stating that the company would invest $80 billion in AI in 2025 alone. For perspective:

For comparison, the James Webb telescope cost $10bn, so Microsoft is spending eight James Webb telescopes in one year just on AI.

For a further comparison, people think the long-in-development ITER fusion reactor will cost between $40bn and $70bn once developed (and it’s shaping up to be a 20-30 year project), so Microsoft is spending more than the sum total of humanity’s biggest fusion bet in one year on AI.

The US’s national defense budget is on the order of ~$850bn, so Microsoft is basically spending ‘a little under a tenth of the annual US military and IC budget’ just on AI.

This is better be worth it…

—//—

Lessons From Red Teaming 100 Generative AI Products ↗

Microsoft’s security research team just published a comprehensive paper on their insights from “red teaming” (*) one hundred generative AI products. The whole report is worth reading (and somewhat sobering):

Lesson 2: You don’t have to compute gradients to break an AI system — As the security adage goes, “real hackers don’t break in, they log in.” The AI security version of this saying might be “real attackers don’t compute gradients, they prompt engineer” as noted by Apruzzese et al. in their study on the gap between adversarial ML research and practice. The study finds that although most adversarial ML research is focused on developing and defending against sophisticated attacks, real-world attackers tend to use much simpler techniques to achieve their objectives.

Lesson 6: Responsible AI harms are pervasive but difficult to measure

Lesson 7: LLMs amplify existing security risks and introduce new ones

Lesson 8: The work of securing AI systems will never be complete

Fun times! 🦹🏼

Link to study.

(*) Red teaming is a security assessment process where authorized experts simulate real-world attacks against an organization's systems, networks, or physical defenses to identify vulnerabilities and test security effectiveness.

---

What We Are Reading

🌳 Creating Healthier Cities Requires a Strong Dose of Nature Research shows that exposure to nature and green spaces significantly reduces health risks from noncommunicable diseases, leading cities and doctors to increasingly treat nature as essential medicine. @Jane

🤔 Perplexity Wants TikTok, but Will It Work? For “just” $50 billion, Perplexity AI wants to merge with TikTok U.S. and new investors, allowing ByteDance stakeholders to retain their equity. @Mafe

🤔 Are Better Models Better? Understanding the difference between “better” and “right” answers is key to identifying the best use cases for Gen AI tools today—and perhaps also evaluating their actual potential for disruptive application tomorrow. @Jeffrey

🚗 Driving Mobility Forward, Vay Brings Advanced Automotive Solutions to Roads With NVIDIA DRIVE AGX Teledriving from the Berlin-based startup Vay has already been live for a while in LA. Teledrivers will provide the car, with the user then driving the car as long as needed. As a median to fully autonomous driving, the company is trying to launch in Belgium in the EU. @Julian

📉 Why America Is in an Alcohol Recession A pandemic trend reversal, along with more moderation overall, has led consumers to significantly reduce their alcohol consumption due to a combination of more mindful spending and increased awareness of its impact on health. @Pedro

🧘 Calm Tech’ Certification ‘Rewards’ Less Distracting Tech After decades of our technology becoming ever more intrusive, our devices are now becoming “calmer” (well, some of them) – it turns out, it might be good business too. @Pascal

---

Some Fun Stuff

👾 A true blast from the past! Play the Atari 2600 classic game Pitfall right in your browser and marvel at what excited us as kids.

🥫 Questions in search of an answer: What are these dots on tin can lids, and the concentric circle relief?

🎯 Why tech might take the Pareto Principle just a tad bit too far.