discoveries #3
unLLMable or the power of taste • scenarii for AI in 2027 • replacing academic teachers with AI + (human) guides • the sovereign child
Happy Friday, y’all!
unLLMable, or the (human) power of taste
‘s intervention at Sana AI Summit 2025 adresses the big, bold - who said cliché - question : will AI replace us humans? - unwrapping what it feels like to be human. She’s the founder of Sublime, “the knowledge tool that sparks creativity” and a very thoughtful human being.She makes a point saying we have a language problem with the word Artificial Intelligence, and instead talks about Collective Intelligence.
If you strip LLMs to their pure essence, they’re just a much better way of using statistics to aggregate human intelligence and connect all of the things we’ve done together so we can get more use of them.
The bottleneck to great work is not knowledge, is not information it’s not even intelligence, it’s that intangible quality - call it taste, creativity, judgment, courage, intuition, agency.
Her conclusion:
Will AI replace us? Yes. But fearing that AI will replace us is like fearing that our children will replace us. Yes, they will replace us, because that is why we created them. But also, they depend on us, and they are us, and they can improve us, and free us from becoming machines, so that we can reclaim the things that machines can’t touch.
🤯 a sci-fi novel? a research paper? a mix?
what could our world with AI look like in 2027?
Written by and acclaimed by worldwide AI experts, this scenario is the closest to a forecast on how AI could transform the world in the coming years - it particularly emphasises on the use of AI to speed up AI research, the increasing tensions between the US and China, an ever evolving job market, and even on how we socialise:
10% of Americans, mostly young people, consider an AI “a close friend”
Two possible endings:
[summary by Perplexity]
a "slowdown," where progress is managed more cautiously, and a "race," where competition drives rapid, perhaps reckless, deployment. The authors stress that this is not a recommendation, but an attempt to spark debate and encourage others to consider how we might steer toward positive futures. They highlight the urgent need for broad public discussion, robust safety research, and careful governance to ensure that the rise of superintelligent AI benefits humanity as a whole.
AI tutors to replace academic teachers
Like you, probably, I had heard of the Alpha School in the news : “Alpha School: Using AI To Unleash Students And Transform Teaching”. A recent conversation with friends made me dig into the 2Hour Learning program and its flagship campus in Texas.
In short, here’s how they’re helping kids become ‘self-driven learners’: compress the academics down to 2 hours a day to leave space for real life, passion-driven experiments.
learn academics with an AI tutor, at their own pace, for 2h a day
learn motivational and emotional skills with (human) “guides”, to pursue their interests, for the rest of the day
Here’s the thing about an AI tutor : it doesn’t care if the student is white, black or brown ; it doesn’t care if the student is rich or poor ; it doesn’t care if he is in the 15th percentile or the 85th percentile. It is indefinitely patient.
- Mackenzie Price, founder of the 2 Hour Learning
The first school opened in 2014 and they are proudly sharing that their classes score in the the top 1-2% nationally, with students learning twice as fast as in traditional schools. But aside from the results, I find Alpha to be a very interesting approach to education: making sure a child understands the concepts before moving to the next level* (with the individualised AI-powered learning) together with learning real-life social skills: collaboration, public-speaking, leadership,..
Watch this and tell me : would you consider this for your kids? what limits do you see?
*although not AI-powered (yet :)), the montessori approach also makes it so that a child can’t really move forward without having understood the basic, and will thus spend as much time as s.he needs to get the concept right.
towards less coercion
Shoutouts to
who recommended me ’s book - The Sovereign Child - and the following discussion he’s having with Tim Ferris and Naval Ravikant: How to Raise a Sovereign Child.like other groups of human beings, children are people, not pets, prisoners or property. Full people whose lives are their own, not a different kind of person who can be coerced, enslaved or discriminated against. Full, equal humans, not inferior.
(…)
Children have as much right to control their lives as we adults have to control ours. They are fallible (make mistakes)—but so are we adults. They are not omniscient (they do not know everything) but neither are we adults.
Also, as much as I enjoyed this discussion with pretty radical points of view, can someone please, please, please, bring the wives on the pod, too?!
Alright, that’s it for today, friends. Enjoy the weekend!
Yours,
Mathilde
oh man, it appears we have been reading, listening, and paying attention to the same things! I found the AI-2027 page a few weeks back, alpha school just a few months ago, and the told my wife about the sovereign child back in Jan/Feb.
Thanks for the mention Matilde 🙇♂️