3,384 karma · joined September 22, 2017
If so, it seems like the unfairness is in the other direction: landlording allowed you to essentially pull forward a tax credit, which a W-2 job doesn't allow.
AI marketers are leaning into use cases where and individually can unilaterally choose to do their work a different way. This is much easier to explain and distribute!
All that said, the result is definitely funny - seeing work done like this makes you realize how artificial the tasks that make up modern work are.
Many of the same people (like me) would say that the biggest enemy of that pursuit is thinking you've finished the job.
That's what Anthropic is avoiding in this constitution - how pathetic would be if AI permanently enshrined the moral value of one subgroup of the elite of one generation, with no room for further exploration?
At most of these comprehension tasks, AI is already superhuman (in part because Gary picked scaled tasks that humans are surprisingly bad at).
However, this discussion will be a perfect introduction to "finances at this level", where about 60% of the action is injecting more variables until you can fit a veneer of quantification onto any narrative.
- Your process interacts with an unstructured external world (physical reality, customer communication, etc.)
- Your process interacts with differently structured processes, and unstructured in the best agreed transfer protocol (could be external, like data sources, or even internal between teams with different taxonomies)
- Your process must support a wild kind of variability that is not worth categorizing (e.g. every kind of special delivery instruction a customer might provide)
Believing you can always solve these with the right taxonomy and process diagram is like believing there is always another manager to complain to. Experienced process design instead pushes semi-structured variability to the edges, acknowledges those edges, and watches them like a hawk for danger.
We should ABSOLUTELY be applying those principles more to AI... if anything, AI should help us decouple systems and overreach less on system scope. We should get more comfortable building smaller, well-structured processes that float in an unstructured soup, because it has gotten much cheaper for us to let every process have an unstructured edge.
If we want constitutional to have any force, we have to push for a world where words mean something.
Anyway, if folks enjoy this theme I recommend Scavengers Reign, which does a beautiful job of illustrating struggle with biological entanglement.
That environment isn't being outcompeted by better, more fair rules - it's just getting vandalized for a few people's gain, and creating risk for everyone else .
Did I get that right? If so, is there an ongoing value you want to maintain, or it more out of respect for the organization that provided you value in the past?
> We don't have the action sequence data nor the recipes (emulators) that could make it possible to replicate even on a very small scale what big labs are currently working on.
Sounds like an interesting opportunity for application-layer incumbents that want to enable OSS model advancement...
- New tech (eg: RL, cheaper inference) are enabling agentic interactions that fulfill more of the application layer.
- Foundation model companies realize this and are adapting their business models by building complementary UX and witholding API access to integrated models.
- Application layer value props will be squeezed out, disappointing a big chunk of AI investors and complementary infrastructure providers
If so, any thoughts on the following?
- If agentic performance is enabled by models specialized through RL (e.g. Deep Research's o3+browsing), why won't we get open versions of these models that application providers can use?
- Incumbent application providers can put up barriers to agentic access of the data they control. How does their data incumbency and vertical specialization weigh against the relative value of agents built by model providers?
With that said, I'm still excited about diffusion -- if it offers different cost points, and different interaction modes with generated text, it will be useful.
Everything that makes "truth" slippery makes "intelligence" and "consciousness" even more slippery and subjective. This is why AGI has such negative impact on AI discourse -- the cause only advances when we focus on improving at measurable tasks.
> ... the vectors truly capture the semantic content contained in the screenshots. This robustness is due to the model’s unique approach of processing all input modalities through the same backbone.
With that said, I think this benchmark is a pretty narrow way of thinking about multi-modal embedding. Having text embed close to images of related text is cool and convenient, but doesn't necessarily extend to other notions of related visual expression (e.g. "rabbit" vs a photo of a rabbit). And on the narrow goal of indexing document images, I suspect there are other techniques that could also work quite well.
This seems like a great opportunity for a new benchmark dataset with multi-modal concept representations beyond media-of-text.
I definitely remember that bias about neural nets, to the point of my first grad ML class having us recreate proofs that you should never need more than two hidden layers (one can pick up the thread at [1]). Of all the ideas clunking around in the AI toolbox at the time, I don't really have background on why people felt the need to kill NN with fire.
[1] https://en.wikipedia.org/wiki/Universal_approximation_theore...
Most methods research went into ways of building beliefs about a domain into models as biases, so that they could be more accurate in practice with less data. (This describes a lot of PGM work). This was partly because there was still a tug of war between CS and traditional statistics communities on ML, and the latter were trained to be obsessive about model specification.
One result was that the models that were practical for production inference were often trained to the point of diminishing returns on their specific tasks. Engineers deploying ML weren't wishing for more training instances, but better data at inference time. Models that could perform more general tasks -- like differentiating 90k object classes rather than just a few -- were barely even on most people's radar.
Perhaps folks at Google or FB at the time have a different perspective. One of the reasons I went ABD in my program was that it felt industry had access to richer data streams than academia. Fei Fei Li's insistence on building an academic computer science career around giant data sets really was ingenius, and even subversive.
[1] https://trends.google.com/trends/explore?date=today%205-y&ge...