HNHacker News
TopNewBestAskShowJobs

MichaelRazum

427 karma · joined August 30, 2019

submissionscomments
MichaelRazum··on China Aims to Use Particle Accelerators to Build Chips and Evade EUV Sanctions
I think this is a bad argument. It's like saying only because someone clever choose path A it is wrong to choose path B.
MichaelRazum··on Meta’s Threads App Won’t Launch in EU on Regulatory Concerns
Crazy. So if meta can't figure that out. How about a small start up that wants to act world wide?
MichaelRazum··on $900k Median Package for Engineers at OpenAI
Do you guys know some good companies that offer interesting remote ML jobs? The package is competitive for sure. I don't care that much about the pay, more about interesting stuff.
MichaelRazum··on Ask HN: How do you deal with large Python code bases?
Hard to tell but that sounds a bit like a test problem. If you have high test coverage (which you must do in a language like python) most bugs should only result through complex interaction.
MichaelRazum··on Germany falls into recession as inflation hits economy
simple, the price is to high. Germany might be "Weltmeister" with one of the highest energy prices...
MichaelRazum··on Ask HN: How do you deal with large Python code bases?
What exactly are your problem's? It is hard to give advice without knowing. To be honest I don't think it makes things easier if you switch to Go, since less verbose.
MichaelRazum··on Meta Open-Sources Computer Vision Foundation Model DINOv2
So what would you do if you want to use it in production? I mean is it worthless? To be honest not sure of what to think about it.
MichaelRazum··on How AI coding companions will change the way developers work
Depends what you are doing and how much code. Especially if you work with new frameworks or apis ChatGPT is extremly helpful. I think especially on small code snippets it is strong. Guess you would have to give an example, what you tried and what you got...
MichaelRazum··on In Battle over A.I., Meta Decides to Give Away Its Crown Jewels
Vicuna looks pretty good. But as said, commercial use not possible.

Why do you think that Llama can be replaced? I mean it is extremely costly to train that thing. And it is there even a clean open source data set for the task?

PS: wouldn't be surprised if Meta, OpenAi, or google will train something for a Billion $ in costs of compute or more.

MichaelRazum··on New proof finds the ‘ultimate instability’ in a solar system model
If you could comment. Is it true that "classic" quantum field theory can't describe chaos, because the operators are linear?
MichaelRazum··on In Battle over A.I., Meta Decides to Give Away Its Crown Jewels
Guess no. That is the trick right? Let's see what people build. If it is good copycat it. If you build something very cool that makes money, be ready being sued by meta...

PS: Not an expert just speculating.

MichaelRazum··on Ask HN: Can someone ELI5 transformers and the “Attention is all we need” paper?
Thanks a lot for sharing. I liked the explanation, although it seems to be not a perfect matching between attention and kernel (as far as I understand). Since as I undestand x_o = Qx and x_i = Qx and y_i = Vx but it doesn't map to Wu dot Wv.

Anyway just curious do you or someone else have more sources in this format?

At least this way it is very helpfull to think about the methods. Kind of agree that the formular does look very very similar to well known methods. But on the other hand, the author doesn't explain the transformers in a similar simple way since it is not obvious why you could stack kernels and get better results.

MichaelRazum··on Ask HN: Those learning about neural networks, what do you find most difficult?
I think, the models (especially transformers) are kind of straight forward.

But the most difficult part is: - intuition behind them, why they work at all. What a good parameter set is - how to check if training goes well - what to do when things go wrong - how to train large models. I mean gpt4 was trained for a long time. For sure they did some things before and during the long training period.

And so on. I mean because the models are rather „simple“, I think the best engineers get paid pretty high, because it is not that much about the models but everything around it.

MichaelRazum··on Google Launches AI Supercomputer Powered by Nvidia H100 GPUs
So can we train now 10t or 100t LLM models? I mean assuming that the dataset is large enough
MichaelRazum··on Google “We have no moat, and neither does OpenAI”
I don't buy it. It is so expensive to train LLM, so that the only hope is to rely on Foundation Models that are open sources by google, msft or facebook.
MichaelRazum··on A morale crisis after multiple Meta layoffs
Facebook is far from done or broke. Especially now where we have the hype and progress around chatgpt. They are in an excellent position regarding that tech.

*if they can execute

MichaelRazum··on Ask HN: What is your longest lasting code?
Life Insurance core code (ANSI C). So contracts closed now have to work for 60 years or more.
MichaelRazum··on OpenAI closes its monster $10B funding round at $27B-29B valuation
It sounds like it is trivial to replicate GPT-4 performance.

I Would be surprised if that is the case.

MichaelRazum··on OpenAI closes its monster $10B funding round at $27B-29B valuation
Agree, somewhere it was mentioned that you get kind of a bond or limited time equity.
MichaelRazum··on HuggingChat
How do this kind of models know when to stop? Does anyone has an idea.
MichaelRazum··on Stanford just released a 386-page report on the state of AI
Not an expert but intuitively, let's say in classical stat's. If you add parameters the danger of overfit increases. So it is far from obvious (at least to me) why it works at all and why bigger is obviously better.
MichaelRazum··on HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace
I asked gpt4 to think through it step by step and then to recheck the answer. At least it got 1/x right. So now kind of impressed.
MichaelRazum··on HuggingGPT: Solving AI tasks with ChatGPT and its friends in HuggingFace
"math genius" not sure about it. I gave it a problem from IMO 2022. In my opinion not a very hard one. It failed even if you give it some hints.

Let R+ denote the set of positive real numbers. Find all functions f : R+ → R+ such that for each x ∈ R+, there is exactly one y ∈ R+ satisfying xf(y) + yf(x) <= 2.

It just struggeld to reason. So I would be very surprised if the plugin somehow helps here.

MichaelRazum··on ChatGPT Plugins
Is it really that hard? I mean ChatGpt is doing the work (that is how I undestand it). Basically if ChatGpt want's to call an external API, it just gives a specific command and waits for the result, then just simply reads the texts and completes the propt. Sounds like a feature that you could prototype in a week of work.
MichaelRazum··on ChatGPT Plugins
The question is how fast can you replicate it?

People will use the best solution. Chrome came after firefox and ie and opera and become more populare because it was better.

MichaelRazum··on ChatGPT Plugins
wow
MichaelRazum··on ChatGPT Plugins
Google=Nokia? It's just crazy that they were leading the field in "AI" and got blown away by OpenAI. Anyway to the expert's in the field, what do you think how hard is it to clone GPT-4 and what would be the hardest part? I had the impression that it is always about compute time and you could kind of catch up very quickly, if you had enough resources.
MichaelRazum··on The everything-is-a-quantum-wave interpretation of quantum physics
I mean the dynamics is driven by a linear operator. But you can observe objects that can't be described this way.

I think Hossenfelder described it in the video the problematic in Chaos: The real problem with quantum mechanics

MichaelRazum··on The everything-is-a-quantum-wave interpretation of quantum physics
For the experts here. As far as I know Quantum Theory is described by a linear differential equation. How can it describe observed non linearity?

The point was made by Sabine Hossenfelder

MichaelRazum··on ChatGPT for Robotics
Very cool! Thanks a lot.
← PreviousPage 3 of 10Next →