wrsh07

11 ms·

181.
by wrsh07 1y ago
In _regarding cocktails_, the note from the author is that it's better to have a cold drink with too much ice than one that isn't staying cold The tacit advice is to drink faster than the ice melts (and if there is no ice, enjoy q
182.
by wrsh07 1y ago
It honestly depends on the headline. I think hn probably has a disproportionate number of haters while Twitter has a disproportionate number of blind believers / hype types. But both have both. Not sure how this compares to YouTube (al
183.
by wrsh07 1y ago
With regard to AI & LLMs Twitter/x is actually the only place with all of the industry people discussing. There are a bunch of great accounts to follow that are only really posting content to x. Karpathy, nearcyan, kalomaze, all of
184.
by wrsh07 1y ago
> Each model was run with the recommended hyperparameters and a maximum token limit of 64,000. No models needs more than this number of tokens I'm a little confused by this. My assumptions (possibly incorrect!): 64k tokens per promp
185.
by wrsh07 1y ago
Yes, but that doesn't mean they aren't finding real value The challenge with the bubble/not bubble framing is the question of long term value. If the labs stopped spending money today, they would recoup their costs. Quickly.
186.
by wrsh07 1y ago
Something I noticed early on when using AI tools was that it was great because I didn't get blocked. Somehow, I always wanted to keep going and always felt like I could keep going. The problem, of course, is that one might thoughtlessl
187.
by wrsh07 1y ago
Add a prompt for referrals that asks them if they're willing to link the discussion that helped them find you! Some users might share it. ChatGPT has so many users it's somewhat mind boggling
188.
by wrsh07 1y ago
This is a bit of a false dichotomy. If it's useful, it's good that it was created. We should be happy about the progress. If it's harmful, we should be happy about guardrails. Friction is a useful tool, I don't begrudge
189.
by wrsh07 1y ago
Would really love to see more functionality built into this. Handling post requests, enabling scripting, etc could all be super powerful
190.
by wrsh07 1y ago
If it's worth using for you in your work, it might be worth your employer striking a deal about data confidentiality so you can use it. But you couldn't use it for work anyway because usage is non commercial. So you need to pay th
191.
by wrsh07 1y ago
To be clear: I agree that opening up model + weights makes it possible for third parties to distill or fine tune If you look at the frenzy of activity that happened after midjourney became accessible, that was awesome for everyone. Midjourn
192.
by wrsh07 1y ago
If it came to light that somebody found a way to use this API in a way that is harmful to society would you be happy that Google could revoke access? Or unhappy? This is a real tradeoff of freedom vs _. I agree that I'm not always a fa
193.
by wrsh07 1y ago
This is a strange take because this is consistent with what Google has been doing for a decade with AI. AlphaGo never had the weights released. Nor has any successor (not muzero, the StarCraft one, the protein folding alphafold, nor any oth
194.
by wrsh07 1y ago
This spoilery link provides context about the op link if that's what you want
195.
by wrsh07 1y ago
There are a few things worth understanding. If someone wants to work at a startup, Google isn't really going to compete for your offer But if you're going to big tech cos you can get them to compete with each other. Know your audi
196.
by wrsh07 1y ago
I've heard of it happening to friends, yes Since any large tech co knows what Facebook offers look like, if you have an unusually good offer they might not believe you without proof and they certainly won't be able to expand the p
197.
by wrsh07 1y ago
Fwiw, I would discourage lying or fraud (which you will inevitably turn to when they ask you for the redacted offer letter including the offer numbers) during an interview process You can strengthen your case extremely well by simply tellin
198.
by wrsh07 1y ago
It was a different era, she was a particularly strong candidate and they had given a low-ball offer. I don't think she had a competing offer fwiw (Less process is normal at smaller orgs, so that's also something to consider)
199.
by wrsh07 1y ago
I expect this is one where if you're coming from an Etsy or Google that gives 5 weeks after 5 years of tenure you could negotiate for preservation of that perk, but it seems like it could be challenging depending on your batna (see def
200.
by wrsh07 1y ago
I actually strongly disagree with this take and 1) after reading the original post immediately linked it to a friend who had an outstanding offer. She got a significant raise from fairly light negotiating 2) personally, I don't think o
201.
by wrsh07 1y ago
Out of curiosity, do you draw the line at other animals? Is it unethical to torture animals? Mammals? If you could understand what whales say to each other would that make you reconsider your position? I tend to not worry much about shrimp
202.
by wrsh07 1y ago
I found the article to be conflating things: > And if a human being is not much more than an algorithm running on meat, one that can be jailbroken and exploited, then it follows that humans themselves will increasingly be treated like th
203.
by wrsh07 1y ago
I expect they don't go out of business: at worst they don't start their next training run quite as aggressively and instead let their new very good model be profitable for a minute Many many companies are currently thrilled to pay
204.
by wrsh07 1y ago
I expect the answer is <$1 as someone who shares a discord server with a friend where we egregiously ping the models
205.
by wrsh07 1y ago
o3 is so good it's worth paying for a minute (just for plus) just to see what it's like I've never used anything like it. I think new Claude is similarly capable
206.
by wrsh07 1y ago
I'm confused by this claim - OpenAI has pretty meaningful revenue. If they monetized free users, they would have even better revenue. The linked post estimates eg $1 per user per month would flip them to profitable.
207.
by wrsh07 1y ago
I think if you use premium models you can run up a bill depending on your plan You can configure it so that you use your API keys, which means you just pay cost but o3 is expensive
208.
by wrsh07 1y ago
If you're parsing JSON or other serialization formats, I expect that can be nontrivial. Yes, it's dominated by the llm call, but serializing is pure CPU. Also, ideally your lightweight client logic can run on a small device/s
209.
by wrsh07 1y ago
Good error messages which are great for humans but also great for LLMs that are trying to debug their initial attempt
210.
by wrsh07 1y ago
The comparison is: for your use case, which is better? Decompress and matrix sum or grab keys and increment individual values? There's some flexibility for how one would do that second increment, but of course naively it might look lik

More ›