kadushka

14 ms·

241.
by kadushka 2y ago
I’ve been using o1, and most recently gpt-4.5, and I haven’t encountered a single case of hallucinations. Could you provide an example of one?
242.
by kadushka 2y ago
I give a similar task when I interview SWE candidates: about half cannot find any bugs (and sometimes see bugs where there are none), despite years of claimed experience in the language/domain.
243.
by kadushka 2y ago
I've been using GPT-4.5 for the last couple of days. To me, it's pretty much AGI already. Or at the very least it is smarter than me.
244.
by kadushka 2y ago
I just tried your prompt: o1, gpt4.5, gemini 2 pro solved it correctly (21-19-36), sonnet3.7 and grok3 failed because of the parsing error you described.
245.
by kadushka 2y ago
Can you provide an example where a top llm (sonnet 3.7, grok-3, o1, gpt-4.5) hallucinated a linear algebra answer?
246.
by kadushka 2y ago
Why take classes if you can learn everything with latest llms? Unless you actually need a formal degree in math ?
247.
by kadushka 2y ago
Here's a good starting point if you want to know more about this topic: https://arxiv.org/abs/2103.00020
248.
by kadushka 2y ago
And thicker
249.
by kadushka 2y ago
That question sounds to me like “have any of you switched away from myspace?” I last used dropbox ~15 years ago. Since then I used S3, Google Drive, Synology NAS, and iCloud. I don’t need to do any syncing across devices: if I need a file I
250.
by kadushka 2y ago
What do you mean? We want images and text to live in the same latent space, and be represented by similar vectors if the two correlate. How else would you want to do it?
251.
by kadushka 2y ago
Not yet. It will probably start later this year: instead of hiring a junior swe, companies will buy gpt-5 or claude-4 subscriptions and ask senior engineers to deliver more. Mass layoffs are still 2-3 years away, but I’m expecting this time
252.
by kadushka 2y ago
I wouldn't be great at finding bugs in it or logical flaws This is what tests are for.
253.
by kadushka 2y ago
So this drug would mess with my head – am I understanding it right?
254.
by kadushka 2y ago
I just asked gpt-4o: “9.9 is larger than 9.11. This is because 9.9 is equivalent to 9.90, and comparing 9.90 to 9.11, it’s clear that 90 is greater than 11 in the decimal place.”
255.
by kadushka 2y ago
Ironically, Soviet Union was the product of the far left ideology.
256.
by kadushka 2y ago
People could tell scaling wasn't working well before the release of GPT 4.5 Who could tell? Who has tried scaling up to this level?
257.
by kadushka 2y ago
Update: finished the paper, and as I suspected, there's a serious downside in speed and memory consumption. LLaDA model has to process the entire output sequence on every time step - without anything like KV cache. Also, full quadratic
258.
by kadushka 2y ago
I'm still not sure I get the difference between 1 and 2. What is "pulls up some associated data into the context" vs ""intentionally"/agentically triggers a lookup for some topic"?
259.
by kadushka 2y ago
As opposed to vector search, or…?
260.
by kadushka 2y ago
I'm still reading the paper, but my main question is how slow is the model compared to LLM of the same size. It seems like to get the best accuracy they need to set number of time steps to the number of tokens to be generated. Does it
261.
by kadushka 2y ago
I have scrollbar in FF on W11
262.
Large Language Diffusion Models(arxiv.org)
7 points by kadushka 2y ago|3 comments
263.
by kadushka 2y ago
Learn llms (book, course, project, etc), practice leetcode, apply for ml engineer jobs. Your background puts you ahead of 99% applicants for those positions.
264.
Zillow built production software without a single engineer(venturebeat.com)
2 points by kadushka 2y ago|2 comments
265.
by kadushka 2y ago
Are we free today? For example, I have to work for a living. If I don't, my family and I will be miserable. Let's just hope that future "men with machines" don't decide to kill the rest of us - I'm not sure w
266.
by kadushka 2y ago
How is learning to program from someone sitting next to you breathing down your neck better than learning to program via a Zoom code pairing session? I’ve been writing software in and out of offices for 20 years, and I honestly don’t unders
267.
by kadushka 2y ago
Do you believe RTO leads to productivity increase? Why? Is there data to back that up?
268.
by kadushka 2y ago
* If you remove the softmax from the attention formula, you end up with linear attention* Sorry, what?
269.
by kadushka 2y ago
I crossed 100k threshold in 2007, few years out of college. Today I make more than 3x that at a startup, fully remote. I could have another 2x if I wanted to work at FAANG.
270.
by kadushka 2y ago
Not a scaling law - a law of physics: how efficiently light can be captured by any optical system. Dragonflies capture ~99.7% of incoming light.

More ›