janalsncm

9 ms·

241.
by janalsncm 3mo ago
The reason is, customer might care a lot more about the prediction than you do, or anyone else does. For example, your customer might really care a lot about some niche prediction like the number of car break-ins in Walmart parking lots .
242.
by janalsncm 4mo ago
For that to be click bait it implies people wouldn’t have clicked on it if any other model were used. IMO the more interesting fact is they ported a game over to iOS.
243.
by janalsncm 4mo ago
Cold comfort I would say. We can only wonder if things would be different if this had happened a year earlier.
244.
by janalsncm 4mo ago
On the healthcare front it’s not even a question. We are already paying via insurance and deductibles and copays etc. W2s don’t see most of the cost because their company pays it.
245.
by janalsncm 4mo ago
In that case, maybe it would make sense for regular people to create their own PACs. Like a privacy PAC. Pay for lobbyists who are more persuasive than our angry emails. As they say, freedom ain’t free.
246.
by janalsncm 4mo ago
When you say “water is wet” and someone asks for evidence, now you will have evidence.
247.
by janalsncm 4mo ago
I read it as a reference to some theoretical limitation like the Halting Problem. Fine, in general it is not possible to tell if code will halt. But if I see a while(1) I’m going to check if the loop can break.
248.
by janalsncm 4mo ago
The purpose of code review is to validate and socialize a change. Machines can get better and better at finding bugs, but until they are hooked up directly to our brains they cannot socialize changes because that involves human understandin
249.
by janalsncm 4mo ago
Note that this article is from April and the ban went into effect July 1.
250.
by janalsncm 4mo ago
Did you benchmark it against simpler methods like BM25?
251.
by janalsncm 4mo ago
That’s a new one. I have never heard of a company which operates like this, giving everyone equal reward no matter how much they contribute individually.
252.
by janalsncm 4mo ago
This is interesting theoretically, but in practical terms it’s hard to apply. RL is already hard. There are many things which can go wrong. You have all of the problems with regular LLM SFT, plus now you have a reward model which can be hac
253.
by janalsncm 4mo ago
It’s possible for something to have diminishing returns. Having a speed limit does not imply the utility of driving is zero.
254.
by janalsncm 4mo ago
I can’t tell if you’re complaining that Meta isn’t saving the whales or that their products aren’t good. If it’s the latter, you should double check their financials.
255.
by janalsncm 4mo ago
Ok, uptime. How do you measure an individual’s contribution to uptime? If Claude goes down does everyone take a hit? If Claude stays up everyone gets rewarded? If so, your metric cannot distinguish between a bad engineer and a good one. If
256.
by janalsncm 4mo ago
1) There is no way to cross reference actual usage because Anthropic won’t tell you how much they think you used. 2) It’s not a nice-to-have, it is a basic accounting question of how much of a product you are getting. 3) Even if it did work
257.
by janalsncm 4mo ago
If products use different amounts of tokens, that’s fine. That’s fundamental to the products. But that’s the numerator, which they aren’t telling you. I’m also saying the denominator isn’t clear. One day it might be a million. Tomorrow 2 mi
258.
by janalsncm 4mo ago
I never strongly considered open weight models before getting throttled all the time from Claude. I bought a 1 year pro subscription but my bet is I won’t renew it. The Claude “usage” UX is very bad. At the most basic level, there is no way
259.
by janalsncm 4mo ago
It’s also worth taking a look at the git history of the repo. The current state is an accumulation of scar tissue, but you can see the actual reasons behind some things by figuring out why they were added in the first place.
260.
by janalsncm 4mo ago
Something doesn’t add up. According to what the article is saying about Meta’s help article, there is no price tier which allows unlimited conversation mode 720 hours/month. Even the highest paid tier can only get 15 hours. But the art
261.
by janalsncm 4mo ago
Do you also believe Terrance Tao was a mediocre mathematician before AI?
262.
by janalsncm 4mo ago
As a rule of thumb, if it is just replacing staging, I don’t have an issue with it. A staged apartment also isn’t “real” so if the AI isn’t adding windows and outlets that don’t exist it’s functionally equivalent to staging. Most real estat
263.
by janalsncm 4mo ago
So in summary, the chance of a person dying from heat or guns is roughly equal in the EU and the US.
264.
by janalsncm 4mo ago
1) It depends entirely on the concept you are trying to prove and how experienced you are in that domain. 2) Not every team will have someone with 20 years of experience in a particular domain eager to spin up a PoC.
265.
by janalsncm 4mo ago
Strongly agree that AI is underspecified. Better name is probably “automation” which at least makes it clear that it depends on what is being automated.
266.
by janalsncm 4mo ago
The fundamental issue is that for almost all students, college is really job training in academic clothing. Universities will never admit this, and some professors delude themselves into believing it’s not true. But it’s the only explanatio
267.
by janalsncm 4mo ago
> Being able to nail a zero-shot greenfield project is relatively easy even for a small model Not really germane to your comment but I hope I don’t sound old when I say I remember a time when spinning up a PoC was a week of work, and a s
268.
by janalsncm 4mo ago
I would respect it if he followed up with “the only real music is dubstep”.
269.
by janalsncm 4mo ago
What is your response to Tidal’s statement > Artificial intelligence and machine learning are not new to music creation, they have just become more commonplace and advanced In other words, there is no bright line. AI techniques have been
270.
by janalsncm 4mo ago
Talent to train high quality LLMs, especially coding LLMs.

More ›