Dawid van Straaten

Generalist, autodidact with interests in mathematics, programming, psychology, and neuroscience. Works with @davidbombal. Opinions are my own!

[Dawid van Straaten @nullaxiom]

@dankrad Apply yourself

[Dawid van Straaten @nullaxiom]

@TracketPacer Someone who studies birds

[BURKOV @burkov]

More thoughts on Jev.

1. It cannot be faster to decision than an LLM and be as good. The time an LLM takes until it prints the first token of the answer is equivalent to the compute spent on making a decision. So, if Jev takes less time to output its full output than it takes for an LLM of equivalent size to print the first token, Jev automatically loses in quality.

2. You can get the same classification speed from an LLM if you do the following:

Let's say you have 3 options you want to evaluate.

1) Prepare 3 separate requests to an LLM, where each request has the same question and only a single option of the three. For example:

- "What color is this dress? Output a single number between 0 and 9 to tell how likely it is that it's RED."

- "What color is this dress? Output a single number between 0 and 9 to tell how likely it is that it's GREEN."

- "What color is this dress? Output a single number between 0 and 9 to tell how likely it is that it's BLUE."

The part right before the COLOR ("What color is this dress? Output a single number between 0 and 9 to tell how likely it is that it's") is the same and is in the KV cache. Then, in parallel, you send only the COLOR word to the LLM and only get from it the very first output token, which will be a number between 0 and 9.

2) Now, you have three numbers, say [2, 3, 9]. You use softmax to convert them to probabilities:

[0.09%, 0.25%, 99.66%]

If you call the LLM with thinking disabled, you get your prediction time as the time of the embedding of the common part of the payload (very fast) + the addition of a single answer option to the KV cache (very fast) + generation of a single-digit output (very fast).

If you enable the thinking, it will take more time to generate the thinking tokens, but the quality will be way better than Jev.

[Dawid van Straaten @nullaxiom]

@CapeTown_Bru Jake is a clown. There should be debuts for the Boks every year. Youngsters coming through and rotation like Rassie is doing. Jake belongs in a museum.

[Dawid van Straaten @nullaxiom]

@hey_madni Looks like what people has been doing with Folds for almost a decade. Amazing! Praise Apple the true innovators

[Dawid van Straaten @nullaxiom]

@TomBukic @fchollet Hilarious love this take 😂

[Dawid van Straaten @nullaxiom]

@NVIDIAGFN Deus Ex

[Dawid van Straaten @nullaxiom]

@kmcnam1 Brilliant 😂😂😂

[John Romero 🤘🏽 @romero]

They're nearly there! Support indie game devs and especially @DoubleFine and @TimOfLegend in their Kickstarter. https://t.co/NpYEeYeChH

[Dawid van Straaten @nullaxiom]

@toneron2 @MIT_CSAIL In the words of Einstein “The secret to creativity is to hide your sources”

[Dawid van Straaten @nullaxiom]

@BBCNews BBC impartial news… according to themselves and no one else. What a joke. Don’t wish Arday’s fate on anyone, but you cannot deny the facts. Cambridge is a joke.

[Dawid van Straaten @nullaxiom]

@FritsKarl @dhh That’s awesome!

[Dawid van Straaten @nullaxiom]

@seclilc @rekdt I’m surprised, got sick from Black Hat last year. But started loading up on vitamins beforehand and ended up okay this year. But think you got into contact with more people than we did.

[Dawid van Straaten @nullaxiom]

Depends on your goals. MacOS imho has gone backwards, and different Linux distros I’ve been using improved a lot. Apple’s hardware in general has a big lead. Would be great to get a fully Nvidia designed SOC (MediaTek does not compete with Apple) on ARM then run Linux on that. Also Nvidia hardware at an affordable price would be amazing…

[Dawid van Straaten @nullaxiom]

@GaryMarcus @TrueAIHound We all know it’ll be here by 2027. Nostradamus’ techbro reincarnation Ray Kurzweil said so with his self-reported perfect prediction score…

[Dawid van Straaten @nullaxiom]

@Gamenadian @DeusExHub I want the original with better graphics. Not Revision.

[Dawid van Straaten @nullaxiom]

@GaryMarcus Hahahahaha my joke prediction for 2027 is on target 🤣 AI companies claiming to solve the Riemann Hypothesis in 2027.

[Dawid van Straaten @nullaxiom]

When you're asked to come up with the Mount Rushmore of rugby... could only come up with one answer.

[Dawid van Straaten @nullaxiom]

@ppp200x @yosp Proton is amazing nowadays. If you install certain Linux distros your FPS is better than on Windows 11 Bloatware.

[Dawid van Straaten @nullaxiom]

@atmoio It still hallucinates. It always will don't fool yourself