LASR

7 ms·

31.
by LASR 2y ago
What you can do with current-gen models, along with RAG, multi-agent & code interpreters, the wall is very much model latency, and not accuracy any more. There are so many interactive experiences that could be made possible at this leve
32.
by LASR 2y ago
So I've done a ton of work in this area. Few learnings I've collected: 1. Lexical search with BM25 alone gives you very relevant results if you can do some work during ingestion time with an LLM. 2. Embeddings work well only when
33.
by LASR 2y ago
Question for the group here: do we honestly feel like we've exhausted the options for delivering value on top of the current generation of LLMs? I lead a team exploring cutting edge LLM applications and end-user features. It's my
34.
by LASR 2y ago
I've been looking for an upgrade from my 2015 5k iMac 27. Might finally pull the trigger on this version. What I will miss is the 27 inch 5k display. Also, my use case is exactly what this iMac is meant for - shared family computer tha
35.
by LASR 2y ago
That's exactly it. I've been peddling my vision of "AI automation" for the last several months to acquaintances of mine in various professional fields. In some cases, even building up prototypes and real-user testing. In
36.
by LASR 2y ago
This is actually a huge deal. As someone building AI SaaS products, I used to have the position that directly integrating with APIs is going to get us most of the way there in terms of complete AI automation. I wanted to take at stab at thi
37.
by LASR 2y ago
I am a fan of 3D printing. And I think you can probably get some circuit traces 3D printed for some niche applications. But active electronics? That's a huge stretch. But more importantly, the economics just doesn't make sense. Co
38.
by LASR 2y ago
The problem with agents is divergence. Very quickly, an ensemble of agents will start doing their own things and it’s impossible to get something that consistently gets to your desired state. There are a whole class of problems that do not
39.
by LASR 2y ago
Yes. Numbers / math is pretty much instant hallucination. But. Try this approach instead: have it generate python code, with print statements before every bit of math it performs. It will write pretty good code, which you then execute
40.
by LASR 2y ago
I lead an applied AI research team where I work - which is a mid-sized public enterprise products company. I've been saying this in my professional circles quite often. We talk about scaling laws, superintelligence, AGI etc. But there
41.
by LASR 2y ago
This is exactly how I feel about FSD in my Model 3. It’s like supervising a teenager learning to drive, ready to take over at anytime. It’s a lot less stressful to just drive myself.
42.
by LASR 2y ago
The only useful part of this might be the mid-air battery-swap tech. A drone with an effectively unlimited runtime would be pretty useful in some narrow applications. But as a pallet? Come on. Cool technical ideas. But terrible applications
43.
by LASR 2y ago
> I’d die of embarrassment and I’d just be sad about wasting my time. Welcome to entrepreneurship! Doing something hard knowing very well that it's most likely is going to be a waste of time is just part of the game. Every once in a
44.
by LASR 2y ago
We've shifted our oncall incident response over to mostly AI at this point. And it works quite well. One of the main reasons why this works well is because we feed the models our incident playbooks and response knowledge bases. These p
45.
by LASR 2y ago
It’s just harder to produce significant gains now. A 10% might be the result of all the chip giants giving it their absolute best. I wouldn’t draw conclusions about the level of competition simply based on the spread in performance. Remembe
46.
by LASR 2y ago
Somewhat off-topic, but it blows me away every time it's brought up that the amount of information you can store in a volume scales by the surface area covering that volume and not the volume itself.
47.
by LASR 2y ago
What do you mean? Digital storage is kind of pointless if we cannot know exactly which bit came from what logical location. ie: it's not like analog audio tape, where we get some signal, but don't really know its exact position.
48.
by LASR 2y ago
With Meta giving out models for free, I wonder what competition even means in this space now.
49.
by LASR 2y ago
I’m pretty certain CS has contracts that limit their liabilities in events like this. Probably a refund is all they’ll be on the hook for. Sadly, damage done like this is just chalked up to an accident, and swept under the rug.
50.
by LASR 2y ago
Signing is meant only to verify the identity of the organization producing the signed artifact. It’s not meant to signify that it’s bug-free.
51.
by LASR 2y ago
I was on a different timezone on holiday, and I got woken up with some panic/anger calls direct from the CEO that we had been hacked. It took several hours of deep-diving, calling up security, backend, frontend engineers until we reali
52.
by LASR 2y ago
You know what I am truly terrified of? When these AI services start keeping memories about your interactions with them to build up a full profile of you. Like chatGPT memories. But then it leaks due to a data breach. It’s not even accurate
53.
by LASR 2y ago
So this opened in my GitHub iOS app at first and I was confused.
54.
by LASR 2y ago
I lead tech interviews for E5/6 level engs where I work. Everyone knows how pointless it is. But it’s become more like paying the tax. For the record, I typically give away the a-ha moment early on and see how the candidate is able to
55.
by LASR 2y ago
Agreed. It's just so much easier to hire for React. It has become the lingua franca of client frameworks. There are still many teams that choose to use something else. But experience in React is most easily transferrable. Even if I had
56.
by LASR 2y ago
Yeah I don’t know about this. Looks like you’re opening yourself up for a TON of liability with some very powerful businesses. But not only that, you’re likely endangering the jobs of the people who take up your offers. Seems icky. I am all
57.
by LASR 2y ago
I continue to hold the strong position that calling LLMs without injecting source truths is pointless. LLMs are exceptionally powerful as a reasoning engine. It’s useless as a source of truths or facts. We have chat bots, chat bots with aut
58.
by LASR 2y ago
We’ve tried. A lot. Custom frameworks and all. There is really no way to make the ensemble behave with an acceptable level of consistency. Where we ended up is now having a frontier model generate a whole tree of possible execution plans, a
59.
by LASR 2y ago
Yeah what was that ad even. I am not a creative. But I do play the piano from time to time. It’s an old 15 year old Roland electric piano. I wouldn’t like to see it crushed. Even if it is obsolete. I bet a lot of actual creatives do have se
60.
by LASR 2y ago
Wow. I was so frustrated with chat that I was almost going to write something like this myself. Now I don't have to :) Curious about the business model here though. How much sales have you had so far, if you don't mind me asking?

More ›