Ahmed Medhat

always hopeful. interested in multi-agent collaboration & coordination tech. previously, graph learning & network science @Meta, neuro ai @CSHL

[Ahmed Medhat @amedhat_]

https://t.co/c23Ygi0s1L

[Ahmed Medhat @amedhat_]

@almostcmb Association between weight, especially stomach fat, and metabolic syndrome (under which all of this falls) is well known.

[Ahmed Medhat @amedhat_]

Your answer is in the question. There are a million sports. There are also a million cognitive tasks, and just like AI models can be good at some more than others (and vice versa), the same is true of people. IQ is equivalent to say that some people are superior in every single sport than others. An absurdity.

[Ahmed Medhat @amedhat_]

if the boy who cried wolf's ultimate true statement is that everyone's drowning, not just himself, and he's the one closest to the waves, then he'd have to have cried a million dinosaurs, not one wolf, for me to stop listening to him. Exactly what have they lied about that makes you confidently ignore a warning this big?

[Ahmed Medhat @amedhat_]

@atmoio Almost the whole knowledge economy runs on people writing on keyboards, speaking to mics, and listening through speakers, which of these is off limits exactly?

[Ahmed Medhat @amedhat_]

@wickedguro @LaxmanP96044 And with $4M+

[Ahmed Medhat @amedhat_]

@pwlot AI capabilities wise or proliferation wise? For physical, achieving the capabilities and having them proliferate can be miles apart.

[Ahmed Medhat @amedhat_]

My bets are (a) on the model side, increasingly complex simulations as there’s not enough counterfactuals in the world once models desire to understand macro effects, and (b) on the ecosystem side, coordination tech and collective intelligence.

Some examples

Models of models that couple alignment with cognitive capabilities. New market policies. A thriving of what so far had been impractical game theoretic market protocols. New philosophies for org charts. Transforming how financial risk is priced via digital twins that establish full audit trails of business transactions, and turn labor into compounding value by do our bidding while we sleep. Email threads becoming market trades. Direct and geometric democracy via digital twins that can finally resolve participation fatigue.

So so many exciting ways of re-defining how the world works.

[Ahmed Medhat @amedhat_]

@weldeiry What’s the obsession with proving that a vaccine rushed through trials is untouchable? almost like it’s holy scripture. But perhaps even worse, as it’s masquerading as objective truth, not belief.

[Ahmed Medhat @amedhat_]

@weldeiry People should also take seriously other possible side effects. I crashed two days after mRNA vaccination, with my cytokine profile going entirely haywire and haven’t recovered from the ensuing chronic fatigue since. There are many such examples.

[Ahmed Medhat @amedhat_]

i don't entirely agree with what he's saying, but your statement is not the dunk it sounds like. problems don't live on a linear scale of complexity. Absent live interaction with our world, brains in a vat will infer quite a few useful things, yet can conceivably hit diminishing returns if subsisting in the vat, which is a good thing actually (by analogy it's similar to how at advent of computing itself we thought all maths was solved but then understood computational intractability). It's not meaningfully different because LLMs are very much the kernel of a modern computational system, and will be beholden to the same fundamental limits.

[Ahmed Medhat @amedhat_]

no it's because ultimately empiricism runs the world, and that requires continuous instrumentation, data collection, and (unless you build the matrix for simulations) human interaction data to derive rules from complex systems (think macro-economic understanding for example by latching onto day to day transactions).

[Ahmed Medhat @amedhat_]

@harryjsisson the power of humor.

[Ahmed Medhat @amedhat_]

@paulg Also, you may or may not be close enough to the wire of this telling anecdote. Many CS students from top schools apply to YC in the summer purely because the reputation is it’s somewhat easier to get into YC than get an internship in a competitive tech company.

[Ahmed Medhat @amedhat_]

Perhaps in terms of VC metrics it hasn’t deteriorated, but both home runs and hit rates have been hurt by cohort expansion.

Especially around Covid when suddenly multiples more companies were incubated by YC, the perceived (and imo the true) quality of a YC alum company has meaningfully declined, and that’s not a particularly controversial experience today (happy to share my personal experience as an angel investor pre and post).

In short, what I’ve seen is that today most strong startups have done the bulk of their raise pre demo day, and the ones that stick around till the end are of measurably less quality on average (ie the claim of “invest in a random portfolio of YC companies and you’ll beat the market”, doesn’t actually hold true when looking at demo day - a formerly core part of the process).

Anecdotally, I would consider the fact that unlike the past, where a large majority of market winners came out of YC, the same cannot be said of the generative AI wave (the only YC one, Replit, is an old company that reinvented itself). Something like Neo, ai grant, or conviction, are much clearer incubation winners when looking at it from that lens.

If YC had rejected (or not attracted to apply) the likes of cognition, pika, any sphere, etc, I’d consider revisiting the evaluation and sourcing processes of YC.

[Ahmed Medhat @amedhat_]

@FabrizioRomano Just another easy to control manager unfortunately.

[Ahmed Medhat @amedhat_]

@arblauvelt I think etymology is a bit like scientific experiments. If the resulting narrative sounds too interesting and curated, it’s usually false.

[Ahmed Medhat @amedhat_]

@hthieblot miRC.

[Ahmed Medhat @amedhat_]

@Bluebirdeyessss @lebanesesami In the UK, it more often than not translates to “you’re an idiot”.

[Ahmed Medhat @amedhat_]

@MichaelAzmy It’s easier to count than using alphabets at this point. Round 8.