zmmmmm

8 ms·

61.
by zmmmmm 2mo ago
I will say here that there are meaningful jobs that especially tech people can engage with. There are gigantic unsolved problems. Climate change, cancer, thousands of rare disorders with no therapy or even good diagnostics. And many of th
62.
by zmmmmm 2mo ago
On top of that, it seems the WFO crowd simply can't help exploiting it to their advantage once they are there. They get less work done, but they hoard higher level decisions about what work should be done. They will say "oh we had
63.
by zmmmmm 2mo ago
it's great but we need a multi-modal model of this quality and price to truly declare victory. But it makes me quite curious, how a text-only model can do so well on ARC-AGI-2 being a set of visual puzzles? It would have to solve it en
64.
by zmmmmm 2mo ago
I started with the Lex Fridman podcast [0] and then I went from there exploring further with reading mostly online material interweaved with discussions with ChatGPT. It is actually a fascinating way to read about history. Admittedly, it i
65.
by zmmmmm 2mo ago
definitely But then it's also true the real arc of these things is only properly visible in retrospect. For example I'm sure historians looking back would include the Iraq war as one of the data points on stupidity even though in
66.
by zmmmmm 2mo ago
Some important context - it appears it applies to submissions from the community but possibly not to core developers: > Contributions in the OpenJDK Community must not include content generated, in part or in full, by large language mode
67.
by zmmmmm 2mo ago
But there are no conclusions. It literally says: > Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis It says a plain prompt is not enough but that is not
68.
by zmmmmm 2mo ago
I read up recently on the history of the fall of the roman empire and it was fascinating how eerily similar it was to the current state of the US. Overwhelmed by corruption and wars they make stupider and stupider decisions on ever larger s
69.
by zmmmmm 2mo ago
The fact that the Chinese models have caught up on benchmarks suggests to me that its likely we will start to transition now into much more of a brand war. It will be subjective qualities that drive our decisions more than measures of absol
70.
by zmmmmm 2mo ago
Muse code is more interesting than the new model in that it seems to natively ship with an orchestrator / subagent pattern built in. Curious if it works well in practice compared to achieving the same thing in Claude code etc?
71.
by zmmmmm 2mo ago
I don't think it's at all certain this won't land back on the same unit economics as the old way. The cost of serving a user doesn't have to be free - it never has been - it just has to not be the dominating factor in yo
72.
by zmmmmm 2mo ago
To me there are different motivations to include imagery and some of them are just fine with AI generation. If the imagery actually goes to the substance of helping to illustrate or explain the actual content then it seems very superficial
73.
by zmmmmm 3mo ago
Sadly while the FLOPS are increasing nicely, total graphics memory is stalled in consumer cards by comparison.
74.
by zmmmmm 3mo ago
There's a growing and fascinating divide between people who see LLMs as more of a "bicycle for the mind" in the vein of Jobs vs those who see them as whollly supplanting the role of human intelligence. I can't help but w
75.
by zmmmmm 3mo ago
The outwardly measurable productivity increase will come but over a time scale of years, and most of that will be in barely measurable increments. Nobody wants to hear that but it's just how the world works.
76.
by zmmmmm 3mo ago
well, the EU just brought down the regulatory hammer which complicates things a bit. we will have to see if it operates as a proxy ban on open models depending on how they go about enforcing all the safety, bias, transparency and data prov
77.
by zmmmmm 3mo ago
I completely agree with you. The question is, are we headed to a situation where it is a new process that still allows the same expression, or is it fundamentally less able to facilitate that expression? we have to get beyond the one-shot s
78.
by zmmmmm 3mo ago
The thing is, it has the same essential property that you don't control it any more. The problem with these tools is the lack of fine grained control means there is no room for you to express your individual creative input. Your total
79.
by zmmmmm 3mo ago
it seems perfect for that - maintaining coherence over 30s-1min type window is achievable and so many ads are built on unrealistic premise to begin with. To me, just reducing the cost alone seems secondary, far more important is putting the
80.
by zmmmmm 3mo ago
The argument is no. We don't / can't trust those because they don't release the weights. How do we verify what they got evaluated is what they actually serve and run. If Anthropic really wants to argue this type existent
81.
by zmmmmm 3mo ago
we should turn this around Private models should be banned because they can't be transparently evaluated. We have to trust the same entities that made them to evaluate them, in spite of their gigantic conflict of interest in doing so.
82.
by zmmmmm 3mo ago
Curious what the fundamental limit on Huawei's capacity is. China has shown if nothing else they know how to scale when they want to. If it came down to just building more of what they know how to do, it would be happening. Is there mo
83.
by zmmmmm 3mo ago
Well, the model that broke out of its sandbox and hacked into huggingface used its own judgement too. If we are going to rely on "judgement" then you have to have a LOT of confidence in that judgement once this hits anything criti
84.
by zmmmmm 3mo ago
It definitely makes me uneasy given the types of behaviours we've been hearing about from these models. Letting them use their judgement can go horribly wrong once a misaligned behaviour is triggered. It loosely translates into a relax
85.
by zmmmmm 3mo ago
Even if you have to deliberately enable a setting and enter a pairing code? It's like saying that the fact the user knows their own password is a vulnerability because they might enter it to log into their account.
86.
by zmmmmm 3mo ago
Pointless anecdote: I asked it to make some slides and it decided to write its own slide rendering engine: > On the format — I dropped reveal.js and wrote a small engine inline instead. Reveal would have meant a CDN load, and a deck tha
87.
by zmmmmm 3mo ago
I'm very happy I can tell when AI wrote something. It keeps everyone honest. I'd be far more concerned if it didn't have a distinct tone and style.
88.
by zmmmmm 3mo ago
It is so rare that a new moving part beats other factors. But the one thing I will say is that there's tremendous advantages to of all things, your queuing system being independent of other architectural pieces : it's the one comp
89.
by zmmmmm 3mo ago
there's a lot more context than that though. I don't think there's any love lost b/w them these days.
90.
by zmmmmm 3mo ago
Can I ask it about DNA without it accusing me of bioterrorism?

More ›