sgt101

52 ms·

391.
by sgt101 2y ago
Yeah - be a little careful there. I've been playing with llama-3.2vision-11b and it can't count more than 4 dots.
392.
by sgt101 2y ago
I've tried with some ppt images rather than Clevr ones and it does much better. It can count circles and triangles and differentiates between them quite well. It can recognise the colours of the objects as well. I think that the faux 3
393.
by sgt101 2y ago
I tested the small model with a few images from Clevr. On first blush I am afraid it didn't do very well at all, it got object counts totally wrong and struggled to identify shapes and colours. Still, it seems to understand what's
394.
by sgt101 2y ago
I'm so out of things ! What do you mean manually computed derivatives?
395.
by sgt101 2y ago
Ok, so what protects you from these folks? What positive measure can be suggested here - that is better than the measures I suggest and subsumes them?
396.
by sgt101 2y ago
mid-60's had neural nets. mid-90's had LeCun telling everyone that big neural nets were the future.
397.
by sgt101 2y ago
It's not quite so - we couldn't handle it, and we didn't have it, so it was a bit of a none question. I started with ML in 1994, I was in a small poor lab - so we didn't have state of the art hardware. On the other hand
398.
by sgt101 2y ago
Here's an example of why I think this would work. GDPR. How many breaches of privacy by large orgnaizations occur in the EU? When they occur, what happens? On the other hand - what's the story in the USA? Alternatively what would
399.
by sgt101 2y ago
well - trusted in the sense that the unwatermarked text was human generated ;o)
400.
by sgt101 2y ago
I think this misses a key point. If there were a law that AI generated text should be watermarked then major corporations would take pains to apply the watermark, because if they didn't then they would be exposed to regulatory and repu
401.
by sgt101 2y ago
You have another one with a network gateway to provide hot failover? Right?
402.
by sgt101 2y ago
Imagine you are running a digital design shop and you want to determine two things: - the software that your users are running is licensed. - the media that is produced is stored on a central resource that's tightly managed and compre
403.
by sgt101 2y ago
Building on this idea people have grounded LLM generated reasoning logic with perceptual information from other networks : https://web.stanford.edu/~joycj/projects/left_neurips_2023
404.
by sgt101 2y ago
Definitely on my mind in the UK. We have maybe 10 days a year when cooling is important, but having sun shades would be really nice for perhaps 50 days a year - so makes more sense than an AC solution here I think. Also they look nice!
405.
by sgt101 2y ago
I love that "system 1" and "system 2" are tossed out as fundamental concepts of mind when they're basically labels for some semi-random ideas based on bullshit studies in a pop sci fluff book. Is there any actual sc
406.
by sgt101 2y ago
I've got my tin foil hat on today! There is a story about how one of Mike Lynchs closest business associates (ex ceo of Darktrace) as been made a minister of state in the UK. Pretty clear signal that she's not to be touched. I don
407.
by sgt101 2y ago
I thought that the Apollo system was barely able to execute on the available hardware, so there was no redundancy. I think that the space shuttle did feature this kind of system though.
408.
by sgt101 2y ago
May be the author could test with realistic data set sizes. Recently I've worked with 26M and 300M data sets. Perhaps something along those lines.
409.
by sgt101 2y ago
What a pity that Brimstone isn't integrated with the Apache.
410.
by sgt101 2y ago
My uncle believed in bible chronology. He believed it so much that he learned to program computers (to calculate and explore patterns in dates) and then built a series of web sites culminating in https://www.biblechronology.com&#
411.
by sgt101 2y ago
I'm a managing consultant, so I don't have much time to do actual work (due to having to chat up clients and tell other people what to do). With cursor I can do what would have taken me an afternoon, previously, in an hour. This i
412.
by sgt101 2y ago
I think Cursor is great, it's really boosted what I can do. I think it's made me 3 -> 5x more valuable. Now, maybe it's done the same for other people as well, but I think my co-efficient was high, and I continue to insist
413.
by sgt101 2y ago
What if you want to make things that are new?
414.
by sgt101 2y ago
We might see that with the inference ASICs later this year I guess?
415.
by sgt101 2y ago
"It may have been the greatest rapid expansion in a given economic sector in British history, and it was the key reason we didn’t experience a Great Depression while Germany, the USA, and France did. " But Great Britain (as it was
416.
by sgt101 2y ago
Super spreadsheet - very useful. Thank you for doing the work and sharing.
417.
by sgt101 2y ago
Well, you make it a capital office, or less draconian... make it an offense punishable by 10 years of hard labour in a dire penal camp somewhere cold. In addition you make it an offense to fail to disclose knowledge of the use of such capab
418.
by sgt101 2y ago
The one fundamental that you have to understand about all government is that it is essentially a monopoly on violence. That monopoly enables things like taxation and broad social consensus and consent. But, it's there, and if Bitcoin b
419.
by sgt101 2y ago
Is there a list of discoveries or siginficant works/constructions made by people collaborating with LLM's? I mean as opposed to specific deep networks like Alphafold or Graphcast?
420.
by sgt101 2y ago
Some say NIMBY's other say water supply. "Water supply issues are already holding back housing development around the city. In its previous draft water resources management plan, Cambridge Water failed to demonstrate that there wa

More ›