woeirua

13 ms·

181.
by woeirua 9mo ago
We must have a different definition of arbitrary. OP ran 2.3 million tests comparing random battles against the original implementation? Which is probably what you or I would do if we were given this task without an LLM.
182.
by woeirua 9mo ago
Had to wait for the models to catch up...
183.
by woeirua 9mo ago
Here’s the thing though, most companies work with small data. The distribution of data set size follows a power law which means that few engineers get to work with petabyte sized datasets. That said, the job market definitely incentivizes p
184.
by woeirua 9mo ago
I’m convinced that there will be people declaring that this isn’t AGI all the way up to the moment that they become permanently unemployed.
185.
by woeirua 9mo ago
People are a bit too eager to jump on the emergence of ads as an indicator that things are slowing down. I view it as the opposite, mostly because of my experience with Codex 5.2 and Opus 4.5. These models are incredible. I think by some de
186.
by woeirua 9mo ago
It's just amazing to me how fast the goal posts are moving. Four years ago, if you had told someone that a LLM would be able to one-shot either of those first two tasks they would've said you're crazy. The tech is moving so
187.
by woeirua 10mo ago
The international order was dead ten years ago. GWB put it on life support with Iraq 2. Obama pulled the plug when he didn’t respond to Putin in Crimea.
188.
by woeirua 10mo ago
Yeah, exactly. After your first 100 cycles you only get 800 miles per charge, which is still way beyond what you need for EVs to replace all ICE cars.
189.
by woeirua 10mo ago
Pumped hydro is not viable in most areas of the world. This is.
190.
by woeirua 11mo ago
They're clearly building better training datasets and doing extensive RL on these benchmarks over time. The out of distribution performance is still awful.
191.
by woeirua 11mo ago
What. Who is buying a $600 camera to take pictures of your stool?
192.
by woeirua 11mo ago
Wait, shouldn't their internal agents be able to do all this work by now?
193.
by woeirua 11mo ago
It really feels like we are determined to simulate every possible task in every possible environment instead of building true intelligence.
194.
by woeirua 11mo ago
Yet another instance of tech bros trying to automate away the things that humans should be doing instead of trying to automate away actually hard or dangerous things that humans should _not_ be doing.
195.
by woeirua 11mo ago
Our family stopped doing gifts after it had devolved into just selecting items off of each person's Amazon wishlist.
196.
by woeirua 11mo ago
Yet another blogpost that looks super impressive, until you get to the bottom and see the charts assessing held out task performance on ASKA and MineDojo and see that it's still a paltry 15% success rate. (Holy misleading chart batman!
197.
by woeirua 1y ago
Is that why they just changed to become a for-profit company?
198.
by woeirua 1y ago
Lack of employee trust in these systems is caused by model (under)performance. There's a HUGE disconnect between the C-suite right now and the people on the ground using these models. Anyone who builds something with the models would
199.
by woeirua 1y ago
We should worry more about doomsday risks that are concrete and present today. Despite the prognostications of the uber wealthy, the emergence of AGI is not guaranteed. It likely will happen at some point, but is that tomorrow or 200 years
200.
by woeirua 1y ago
There's this weird disconnect in tech circles, where everyone is deathly afraid of AGI, but totally asleep on the very real possibility of thermonuclear war breaking out in Europe or Asia over the next 10 years. There's already
201.
by woeirua 1y ago
300k lines of AIslop that probably would've been 20k lines of code from a human. And the human has no ability to maintain it, or to reason about it.
202.
by woeirua 1y ago
I'm sure they all know it's happening. But the incentives are all misaligned. They get promotions and raises for pushing the frontier which means showing SOTA performance on benchmarks.
203.
by woeirua 1y ago
There's a huge disconnect between what the benchmarks are showing and what the day-to-day experience of those of us using LLMs are experiencing. According to SWE-bench, I should be able to outsource a lot of tasks to LLMs by now. But
204.
by woeirua 1y ago
Gold has historically underperformed VTI badly. This might just be a small blip. Who knows.
205.
by woeirua 1y ago
We'll know next year if it's not the right move. If we get another wave of incremental upgrades at best then AI will clearly be in a bubble and it's time to look for the exits. If we keep getting major advances then any coun
206.
by woeirua 1y ago
We have to acknowledge that most companies are horrible at recognizing a bad hire and then appropriately handling the situation. A lot of companies profess to "hire fast, fire fast" but very few actually do. Until that changes,
207.
by woeirua 1y ago
Because search fundamentally requires curation. An algorithm has to "decide" which videos are most relevant to you. Otherwise, you'd be flooded with irrelevant results every time you make a search query.
208.
by woeirua 1y ago
Exactly, they shouldn't be the ministry of truth. They should present balanced viewpoints on both sides of controversial subjects. But that's not what they're doing right now. If you watch too many videos on one side of a sub
209.
by woeirua 1y ago
I might agree that the algos making recommendations on the sidebar might not matter much, but the algos that control which videos show up when you search for videos on Google, and also in YouTube search absolutely do matter.
210.
by woeirua 1y ago
It seems to me that a lot of people are missing the forest for the trees on misinformation and censorship. IMO, a single YouTube channel promoting misinformation, about Covid or anything else, is not a huge problem, even if it has millions

More ›