dongping

The robotaxi business model is the total opposite of scaling. At my previous employer we were solving the problem "block by block, city by city", , and I can only assume that you are living in the right city/block where they are tackling.

If there's a high premium, then one might want to wait for a year or two for the premium to vanish.

I, for one, don't like American canned comic books or IPs produced by faceless corporations and prefer artisanal Japanese manga.

Not sure about better models, but DALL-E3 still seems to be having problems with hands:

https://www.reddit.com/r/dalle2/comments/1afhemf/is_it_possi...

https://www.reddit.com/r/dalle2/comments/1cdks71/a_hand_with...

On the other hand, it is very difficult to distinguish between "emergent behavior" and "somehow leaked into our large training set" for LLMs.

Aside from the mandatory negotiation with the employee co-determination (works council), the layoffs "must be determined according to the principles of social selection [...] (t)his is often perceived as a major obstacle [...] it would be disastrous to have to dismiss top performers simply because they are younger than other employees or have been recently hired."

[Source] Random link from the Internet so that I don't need to translate: https://www.emplawyers-muenchen.de/wp-content/uploads/files/...

The new bzlmod mechanism from Bazel can be count as a kind of package manager.

Bazel streamlines the build process with native code (C++/Rust) in Python, making the experience a breeze. However, the drawback is that you would need to transition to Bazel.

Maybe some links please? I also don't want to share a fabricated story.

To commoditize your complements, so that you don't have to pay license fee to your suppliers.

In a way, the German OEMs have been trying to do so, but mostly via different standardization efforts (OpenSCENARIO for example) so that they can easily change suppliers.

Bertrandt, IAV, EDAG, to name a few (link in German):

https://de.wikipedia.org/wiki/Entwicklungsdienstleister#Top_...

Not that I don't agree. I would however rank Sci-Hub, Library Genesis and Anna's Archive higher than Wikipedia. And don't forget about Internet Archive and Arxiv.

This vast amount of high quality human knowledge was not available in the good old internet 20 years ago.

Actually, that's great in my opinion.

Assuming that semi-convincing misinformation spreads everywhere, people will finally have to find the original source of a certain statement, verify their "knowledge supply chain", and maybe use logic to evaluate every single statement made.

This case was presented to the House Committee on Education & the Workforce in the hearing on "How SCOTUS’s Decision on Race-Based Admissions is Shaping University Policies":

https://www.youtube.com/watch?v=4Zu5cdfv9kk&t=2587s

At least this specific case was cited in a congressional hearing:

https://www.youtube.com/watch?v=4Zu5cdfv9kk&t=2587s

The discussion related to device attestation: https://github.com/antifraudcg/meetings/blob/837baaee953bc76...

The rendered version (using URL Fragment Text Directives): https://github.com/antifraudcg/meetings/blob/main/tpac/tpac-...

Perhaps serving plain text data is cheap enough so that it can be done as a non-profit/hobbyist project?

The problem is that people don't like to read text, and thus the above proposed social network might not have widespread adoption, thus reducing the network effect.

Thanks for testing it. I canceled my ChatGPT Plus a few months ago (when they changed the color from black to purple IIRC).

So I only tested that with the GPT-4 API, with the following results:

{'id': 'chatcmpl-7uVF5xGqR1oEzXITw3WZYsnB4Yzt8', 'object': 'chat.completion', 'created': 1693700819, 'model': 'gpt-4-0613', 'choices': [{'index': 0, 'message': {'role': 'assistant', 'content': '353538'}, 'finish_reason': 'stop'}], 'usage': {'prompt_tokens': 11, 'completion_tokens': 2, 'total_tokens': 13}}

This was the response that I got from the GPT-4 API yesterday:

{'id': 'chatcmpl-7uVF5xGqR1oEzXITw3WZYsnB4Yzt8', 'object': 'chat.completion', 'created': 1693700819, 'model': 'gpt-4-0613', 'choices': [{'index': 0, 'message': {'role': 'assistant', 'content': '353538'}, 'finish_reason': 'stop'}], 'usage': {'prompt_tokens': 11, 'completion_tokens': 2, 'total_tokens': 13}}

Maybe they fine-tuned the ChatGPT version better, or fed it to an calculator.

One can resort to traditional vertical multiplication (which requires patience), or do

897*394 = (900-3) * (400-6) = 900*400 - 6*900 - 400*3 + 3*6 = 360,000 - (5,400 + 1,200) + 18 = 360,018 - 6,600 = 353,418

I'm not sure about its math, but GPT-4 fails miserably at simple arithmetic questions like 897*394=?

The GPT-3.5 turbo is fined-tuned for arithmetic according to ClosedAI (noted in one of the change logs), so it is sometimes slightly better, but nevertheless always fails equations like 4897*394=?

The crypto scams highlight the consequences of lacking regulation. Incorporating lessons from a century of financial turmoil could make crypto viable (implementing laws and regulations as code), though potentially less attractive to those seeking an alternative.

I'm not sure if it is meaningful to divert the topic to an analogy that is never precise. But I think we always rerun the same code (equivalent to the same procedure in the papers) in the CI or your workstation to debug. Replicating the result of the program doesn't mean that one would have to rewrite the code, as in replicating the result of a computer vision paper may only include code review and running the code from the paper.

In the sense of replicating the results, we do have CI servers and even fuzzers running for our "code replication".

Then those unconfirmed results are better put on arxiv, instead of being used to evaluate the performance of scientists. Tenure and grant committees should only consider replicated work.

In some fields, replication is already the prerequisite to benchmark the SoTA. So the incentives boil down to publishing them along with negative results. Or as some have suggested, make it mandatory for PHD candidates to replicate.

Though, it seems that it is possible to game the system, by creating positive/negative replication intentionally, to collude with/harm the author.

While it is a lot of work, I tend to think that one can then always publish preprints if they can't wait for the replication. I don't understand why a published paper should count as an achievement (against tenure or funding) at all before the work is replicated. The current model just creates perverse incentives to encourage lying, P-hacking, and cherry-picking. This would at least work for fields like machine learning.

This is, of course, a naive proposal without too much thought into it. But I was wondering what I would have missed here.

"peer code testing" is already the job of the CI server. As it is nothing new, it probably is not going to catch on.

It already fails to answer rather simple (but long) multiplication like 975 * 538, even if you tell it do it in a step-by-step manner.