The 2.5% Revolution
AI agents still can’t finish real work – yet ChatGPT is already thinning our neural wiring, remote work (not AI) may be the thing gutting entry-level jobs, and the whole datacenter boom is stuck waiti
Dear Friend,
I am currently on the road for a series of client sessions – as part of this, I had the chance to sit in on a full day of AI sessions taught by some of the leading experts in the field at a top university – all geared toward a business audience. What stood out to me was less the specific content, but rather the fact that the sessions were all very, very similar. Here you have three experts talking about AI – and they all say the same thing. This might be indicative of the fact that we all still know very little about how AI truly will play out in the real world – hence we all make the same claims, use the same examples, and come to the same conclusions…
And now, this…
Headlines from the Future
Maybe Remote Work Is the Culprit for Young People’s Hiring Issues. You have seen the headlines – young professionals increasingly have a hard time finding jobs and AI is to blame. The argument is easy to follow and certainly makes sense at first blush. A newly published paper begs to differ – what if it’s not AI (which is going through its own identity crisis at the moment, trying to prove its ROI), but remote work? The argument goes like this:
Early-career workers require more supervision than experienced hires, and build important skills, knowledge and social capital by observing and working alongside senior colleagues. Working from home adds friction to these processes, making entry-level workers more costly to bring on board in terms of time and resources and slowing their prospects for promotion. As such, the rise of remote work has worsened the trade-off for hiring entry-level workers, while leaving the calculus for senior hires unchanged.
If this proves to be true, you can expect a double-whammy hitting young professionals – as AI surely will have an additional effect on their job prospects.
━━━━━
AI Agents (Still) Suck. Scale Labs just updated their Remote Labor Index (RLI) – a measure of how well AI agents are actually able to do work in the real world (“Evaluating the capability of AI agents to perform real-world, economically valuable remote work”). The tl;dr:
Absolute Automation is Near Zero: Current agents perform near the floor. At the time this leaderboard was launched, the highest-performing agent (Manus) achieved a 2.5% automation rate, with other models performing worse. This indicates systems fail to complete the vast majority of projects to a professional, client-ready standard.
The upshot? They are getting better.
━━━━━
Your Brain on ChatGPT. A new study from MIT’s Media Lab shows the correlation of tool use (in three groups: ChatGPT, Google Search, and no tool use) with brain activity. With the caveat that this is a small study, the results are not pretty:
EEG analysis presented robust evidence that LLM, Search Engine and Brain-only groups had significantly different neural connectivity patterns, reflecting divergent cognitive strategies. Brain connectivity systematically scaled down with the amount of external support: the Brain‑only group exhibited the strongest, widest‑ranging networks, Search Engine group showed intermediate engagement, and LLM assistance elicited the weakest overall coupling.
As the study concludes: “We demonstrate the pressing matter of exploring a possible decrease in learning skills.” Ouch.
━━━━━
Energy? Water? RAM? GPUs? No, Electricians Are the AI Datacenter Bottleneck. Add this to your AI datacenter bingo card: not only are GPUs and RAM in short supply, energy is a major limiting factor, and water consumption is a massive concern; we also don’t have enough electricians to build out the infrastructure.
That said, the Lake Mariner site is about one hour away from the Buffalo Bills’ stadium, which delivers another benefit. The completion of a massive refurb at the Bills has freed up hundreds of electrical contractors. And it is trades, specifically electricians, which are the biggest bottle neck for datacenter projects, said Farrell.
What We Are Reading
‘An Equal and Habitable World Is Possible’: Academics Set out Sweeping Vision for Planetary Survival A rare piece of good news! A sweeping new report makes the case that a fairer, greener world is materially possible and lays out exactly how to get there. @Jane
Hackers Hijacked Instagram Accounts by Tricking Meta AI Support Chatbot into Granting Access Hackers exploited Meta’s AI support chatbot to hijack Instagram accounts by simply asking it to reset passwords without ever needing access to the victim’s email. @Mafe
AI Has Ruined the Job Market AI tools have rendered the old signals meaningless. In their place: a chaotic hall of mirrors where applicants and recruiters are both feeling lost. @Jeffrey
The New Ivies Over the weekend, my 17-year-old niece told me she’s never used AI because her school’s stance has basically been: “Don’t use it; it’s cheating.” That honestly alarmed me a bit, because the gap between how schools are treating AI and how employers expect people to work with it is growing fast. @Kacee
The Metaverse Fever Dream Phenomenal long-form exploration of the shitshow that is the “Metaverse.” If you ever asked yourself “what was THAT all about?” – here is your answer. @Pascal
Down the Rabbit Hole
😱 Another hard-hitting AI critique from Ed Zitron. Disagree with him, but do yourself the favor of reading it – it’s important to see all sides of the discussion: AI doesn’t have ROI.
🦹 AIs are getting pretty good at working with our day-to-day tools – which also means they are becoming powerful vectors to attack us: ChatGPT for Google Sheets exfiltrates workbooks.
🪖 First we had fitness app Strava leaking the secret locations of US airbases; now we have ad-tracking tech exposing the secret locations of US troops: The ad-tracking industry is exposing US soldiers on the battlefield.
🧑⚖️ This is a little bonkers and shows you how far AI has come (in certain domains): In a rigorous blind study, law professors overwhelmingly preferred AI-generated answers to student legal questions over answers written by fellow law professors – and flagged the AI answers as potentially misleading or harmful far less often.
💰 More and more companies are putting the brakes on their lavish AI (token) spending – Uber just instituted a (sensible) $1,500 monthly cap per tool. Simon Willison has a thoughtful back-of-the-envelope calculation of what that means in real terms – and what the implications are.
↗ Dive into the deep end: Access our complete collection of 2,800+ radical links.
Should We Work Together?
Hi! I’m Pascal from radical. This newsletter is our labor of love. When we’re not writing, we run radical, a firm that helps organizations navigate the future without the “innovation theater.” Most leaders want to seize new opportunities, but they hate endless strategy decks that go nowhere. At radical, we don’t run “projects”; we build your organization’s internal capacity to handle disruption and change. Our goal is to make you future-proof so you can stop reacting to the world and start shaping it. If you’re interested, let’s jump on a call to see if we’re a good fit. Click here to speak with us.