golol

15 ms·

211.
by golol 2y ago
Yea at some pont the style and tone of these assistants needs to be seriously changed, I can imagine a lot of their RLHF and instruct processes emphasize sounding good vs being good too much.
212.
by golol 2y ago
To me it looked they used one command that did copy+paste into ChatGPT both.
213.
by golol 2y ago
What Openai has done time and time again is completely change the landscape when the competitors have caught up and everyone thinks their lead is gone. They made image generation a thing. When GPT-3 became outdated they released ChatGPT. In
214.
by golol 2y ago
Yes and that either works perfectly and Stackoverflow is not needed, or it doesn’t work and we keep using Stackoverflow. In reality it is of course a mix of the two. My point is, we’re not just going to sit around and twiddle our thumbs if
215.
by golol 2y ago
I personally hope that nothing changes in this regard. I think it’s great if everyone can use the content that was voluntarily posted online for the whole world to see.
216.
by golol 2y ago
Who is harmed by this deal between Stackoverflow and OpenAI? What’s the problem with OpenAI profiting of the content on Stackoverflow? I profit of it, everyone does. The whole point is to share knowledge for anyone to profit off. There is a
217.
by golol 2y ago
Stackoverflow is and will be fine, what are you basing this panic on?
218.
by golol 2y ago
This is not true. If the right abstractions are used, proofs in Lean for example can be quite readable and concise.
219.
by golol 2y ago
> I don't actually understand why this is, as it's not like SpaceX is relying on any sort of just extremely well guarded secrets It's good management.
220.
by golol 2y ago
Yes. Right now there are probably a dozen planet scale disasters brewing. We’ll just have to manage them, like we did before. That’s sort of the price with technology, you solve your problems and get new problems, but along the way life get
221.
by golol 2y ago
If we can influence it we can hopefully also control it
222.
by golol 2y ago
My personal belief is that translating an informal proof from a paper to a proof in Lean, given that the prerequisites are already formalized, is an O(n) operation where n is the length of the proof. It's just translation. Right now th
223.
by golol 2y ago
It is a fundamentally different manner. Proofs in Lean are formal all the way down. The kernel checks the validity of a proof. This is a finite program checking the validity of the kernel to high degree once is different to checking the val
224.
by golol 3y ago
As long as all participants are well-informed then there is absolutely no ethical issue...
225.
by golol 3y ago
The nature of computation is just not about extendability of memory for me. It's just a technicality.
226.
by golol 3y ago
People always makes these points of things not being technically Turing machines because they are finite, they have finite memory. Well, obviously! Your computer has finite memory too. It doesn't matter in practice, and the conceptual
227.
by golol 3y ago
Haha its the funny day where we all post unfunny misinformation
228.
by golol 3y ago
Did you not have a physics class around the same time you learned calculus and linear algebra? That makes it obvious what the application is.
229.
by golol 3y ago
I feel like you must be a using a wrong approach to searching for mathematics. In your case I would search for a script or book with title "Introduction to probabilify theory" or "Introduction to mathematical statistics"
230.
by golol 3y ago
> Consider "modern" to mean NN/connectionist vs GOFAI AI attempts like CYC or SOAR. Well this is what Im trying to say too!
231.
by golol 3y ago
Well transformers still have some plausible competition in NLP but besides that, there are other fields of AI where convnets or RNNs still make a lot of sense.
232.
by golol 3y ago
No, but it is weird to use "modern" here. Modern suggest a longer timeframe. I would say machine learning deep NNs is modern AI. It's just not true that everything not transformer is outdated, but it is "kinda" true
233.
by golol 3y ago
>The latest Falcon 9 revision is ~$67m per launch And that is the price, the cost is unknown but likely lower.
234.
by golol 3y ago
Ok so they could leave Starship in orbit and launch payloads like that. The idea however is to launch payloads WHILE testing reentry and landing. This requires an engine relight in orbit.
235.
by golol 3y ago
One technical modification: They need engine relight in orbit to work to deliver payloads, otherwise Starship will stay in orbit and they can not test reentry.
236.
by golol 3y ago
I don't understand this negative sentiment. We are finally making progress on real physical robot labor. This has nothing to do with "imaginary number go up", this is about more labor being available to assemble, clean, trans
237.
by golol 3y ago
So AI = AGI, I see.
238.
by golol 3y ago
No nothing about personal perception. It is completely normal and common that someone profits off me while I myself am also happy with the situation. It is called a trade. Every interaction that's not a zero-sum game is two people prof
239.
by golol 3y ago
Yea of course this is just an LLM interacting through a very crude interface with some control algorithms, but I think it is amazing that we sort of have an approach for both ends of the complexity spectrum now: LLMs for high level, vague,
240.
by golol 3y ago
> If someone profits off me it must mean I am being exploited. This is a fallacy.

More ›