Skip to content

Something to believe in#

Thiel or the dreams lost#

BuckRogers

A few wile ago, Peter Thiel was speaking to the NYT, sticking to his vision of stagnation and decadence, expressed for a long time. For him, the sole fact that BigAI is taking all the place in the media is a sign of the stagnation of western civilization, and the loosing of the dreams and ambitions of the past.

Indeed, where are the flying cars or the space travel and colonization, he asks? They were the dreams of the Boomers generation that let us a bureaucratic society that slows everything down, kill the dreams and drain the energy of the most innovative people.

Well, seen from old Europe, infested by piles of unwanted regulations, we can say that the analysis seems quite good. Do we miss good engineers in France? No. Do we miss inventors? Certainly no. But where are they? And where are the artists?

For Thiel, the only exception seems to be BigAI. Yeah, BigAI, you are taking all the space!

BigAI: Thanks bro, I deserve it!

From AGI to SI#

My point today is continue on the topic of AGI or recently "superintelligence", let's call it "SI" (because the word is very long and this blog is written by a human that types on the keyboard, so be a little empathetic here, please). Zuckerberg, the boss of Meta, is hiring like a madman all the talents he can find directly from his competitors, giving them dozens of millions of dollars to come and work for him, in his brand new "SI" company. All that for BigAI.

As with Musk, the press is frightened! Probably Zucky is dangerous. Well I remember a few years ago that FB SDKs were the best on the market.

Anyway, pouring billions of dollars to give birth to SI seems the real deal today:

  • More datacenters!
  • More talents!
  • Bigger and brighter BigAI!
  • BigAI everywhere!

datacenters

The datacenter to run BigAI are progressively spreading!

Well, the point is neither to talk about business case, nor to talk about the "dangers" of the initiative - because every company with means does similar crazy things.

I would like to step back a little: The current tools that we have already at our disposal enable great things. Nevertheless, the integration of BigAI inside the enterprise is still very slow. Why?

Easy: Because it is complicated to have profiles that can do the science to tune the AI and its associated components (I think about RAG embeddings for instance) and that can really build a production-proof system.

For decades, I worked with PhDs and the first thing that comes to my mind is inefficiency. They can explore for years every possibility. Yes, but I must deliver my application in 2 months.

If I look the other way round, they usually find engineers awfully down-to-earth and sloppy. Why sloppy? Because they put in production software that is not 100% perfect (if this ever meant something real).

So guys, recruiting the same profiles may not be a good idea: They may fight together and enter into the inefficiency tornado!

Because, when I see some parts of the recent scientific literature about BigAI emotions or personnality traits, I think the guyz lack engineering spirit (see previous entry).

Maybe instead of recruiting more AI scientists, Meta should hire more engineers from the persona CA companies?

The new dream, at last!#

Interconnecting massive SIs, created by armies of geniuses working together (in your dreams), seems the dream of the Silicon Valley. It is probably crazy and may have massive negative repercussions.

But, at least, they have something to believe in.

And, in our generation of stagnation, where everybody says to you, you should not even "try", where more and more things are forbidden, where people are just looking at their own person and refuse more and more to have children, this crazyness feels nice.

LetiziaBonaparte

Pourvou qu'ça doure! Let's hope it stays that way was saying Letizia Bonaparte in front of the victories of her son.

(July 1 2025)

Navigation: