anu7df

6 ms·

1.
by anu7df 3d ago
For me opus-4.8 was the most useful coding model. Sure fable is better at planning but is expensive to use as a daily driver. But even fable, when it fails, fails in such strange ways that I am now convinced this intelligence is an illusion
2.
by anu7df 4d ago
Cracks are starting to appear. Open Ai and Anthropic are publicly asking for slowdown in AI research. Translation: We see this technology not being any more useful than what it is now, no AGI is coming, and the first one to accept this and
3.
by anu7df 6d ago
Everyone is rushing to send signals of AI growth to the market. Their continued valuation growth depends on it. There is no real use story that cannot be met by smaller distilled models which are far more energy efficient. I am convinced th
4.
Lamb Grazing in Solar Farms(spacedaily.com)
5 points by anu7df 17d ago|0 comments
5.
by anu7df 7mo ago
Add a nice prior from a photograph, combine multiple measurements and presto you can pick out a person from a crowd with reasonable accuracy. No?
6.
by anu7df 7mo ago
Sci-fi ? You mean historical fiction!
7.
by anu7df 8mo ago
I understand model output put back into training would be an issue, but if model output is guided by multiple prompts and edited by the author to his/her liking wouldn't that at least be marginally useful?
8.
by anu7df 8mo ago
Does coffee accelerate rusting?
9.
by anu7df 8mo ago
You are ignoring the obvious difference between errors introduced while translating one near-formal-intent-clear language to another as opposed to ambiguous-natural-language to code done through a non-deterministic intermediary. At some poi
10.
by anu7df 8mo ago
This is pretty much how I use LLMs as well. These interactions have convinced me that while the LLMs are very convincing with persuasive arguments, they are wrong often on things I am good at; so much so that I would have a hard time openin
11.
by anu7df 8mo ago
I have nothing against automated code completion on steroids or agents. What I cannot condone is not reading and understanding the generated code. If you have not understood your agent generated code, you will be "surprised" for s
12.
by anu7df 8mo ago
I don't know about you, but when the creator of a software says I have not read any of the code, I don't want to install or use it. Call me old fashioned. Really hoping this terrifying vibe coding future dies an early death before
13.
by anu7df 1y ago
Also, if you do not want to spend on any specialized fountain pen flush solution, diluted windex works very well. From the smell I am willing to bet that most fountain pen cleaning solutions are exactly that.
14.
by anu7df 1y ago
Re-entry permit is granted for 2 years and the only requirement is that you apply for it while in country. Usual processing time nowadays is around a year so in effect you need to apply once every 3 years for it. They have the option of sen
15.
by anu7df 2y ago
Oh great.. This looks exactly like some PhD advisors I've heard of. Creating a list of "ideas" and having their lab monkey PhD students work on it. Surefire way to kill joy of discovery and passion for science. Nice going goo
16.
by anu7df 2y ago
May be I am an aberration, but I actually seek out coffee shops for good coffee. A nice espresso, cappuccino and on occasion black drip or pour over. I will never be as comfortable at a coffee place as I am at home, so why bother. But yes,
17.
by anu7df 2y ago
I don't think this is a hypothetical use case, at least for me. I like writing on paper with a fountain pen. But would like a digital version of the notes that are searchable. Reasonable ocr exists for conversion to text, but this woul
18.
by anu7df 2y ago
Check out textify on appstore. I have found the accuracy to be great even for my terrible hand written text.
19.
by anu7df 2y ago
The only question I have is whether Goldman is shorting NVIDIA..
20.
by anu7df 2y ago
I am sure you are already doing this and the "what their pain point is" phrasing of this post is only for succinctness. In my experience, asking a practitioner about their pain point is seldom the starting point for an unbiased co
21.
by anu7df 2y ago
Official term should be AI 2.0 double the artificial, zero intelligence
22.
by anu7df 3y ago
I really believe this "application" is the result of thinking about tests as a chore and requirement without great benefits. Your thought of LLM writing application give the tests is interesting also from test pass/fail as op
23.
by anu7df 3y ago
I keep hearing this all the time. May be I am (and a whole generation of me) an aberration, but I have no plans to switch to a car on demand. Let me be clear that I love public transport, like a metro. Something about the small confined spa
24.
by anu7df 3y ago
I don't think in this use case that would be very logical. If we want to deploy survey telescopes and use SpaceX for it, what is the harm. Once deployed it is deployed. On the other hand if it was about deploying the nuclear deflection
25.
by anu7df 3y ago
Good god. Anyone else who is curious about the human centipede, please don't Google it. Trust me on this one.
26.
by anu7df 3y ago
The origins of the Singularity notion is interesting. But I find it amusing that we could think Singularity is near (mostly) on the basis of some LLMs. Unless we can create a true AGI (It is difficult to precisely define that true Scotsman,
27.
by anu7df 3y ago
I really do not understand why most people seem to think deficit spending is not the cause of inflation. Print a bunch of money and make a lot of people millionaires either on paper or realized gains. Why would their attempt to acquire anyt
28.
by anu7df 3y ago
Office space reference.
29.
by anu7df 3y ago
Not exactly sure if this would qualify as an LLM in the GPT4 sense. But for no hallucination this seems good: https://www.thirdai.com/pocketllm/ Full disclosure. I know the founder, but not really associated with the c
30.
by anu7df 3y ago
I thought concussion was from the brain sloshing around inside the skull . How does a tongue "set-belt" around the skull help with that?

More ›