SgtBastard

Your industrial activity is authoring software, not constructing buildings.

I thought you may have been nitpicking previously but you’re just incorrect.

The authoring of software is o absolutely an industry. The nitpick seems odd.

Lexus has the quality of Toyota behind it, so the analogy doesn’t quite hold.

What has the veneer of sophistication but is actually garbage quality…

Ferrari vs Alfa Romeo, perhaps?

As someone with 100s of hours in SupCom and 1000s in SC2:

SupCom is interesting to play and boring to watch. Tier 1 rushes are map specific to your point… but then Tier 1 units are irrelevant.

Risk would be similarly boring to watch. It’s not that it’s a bad game (quite the contrary).

Blizzard has to balance the meta as the meta is so extensively studied - because it has an audience.

> Not that Starcraft plays bad, but none of the fluff contributes to gameplay. And that gameplay takes a lot more micromanagement than it needs to(some people like this. me, not so much)

As the gen Zs would say: Skill Issue.

Authentication is (proving) who you are.

Authorization is the process of determining if someone can do something.

That you’re conflating multiple distinct concerns doesn’t build confidence.

Temperature = 0 would always return the highest probability token at every generation step and given a single distribution of tokens at each generation steps,that would become deterministic.

Where that gets murky is… perturbation inside the models themselves, such as Mixture of Experts (MoE) models who’s more internal parameter activation is not so deterministic (and therefore the tokens they generate are not deterministic across runs)

Are you asking “is the probabilistic inference model 100% deterministic?”

If so: no.

Just leak them to the inference providers, obviously /s

If you have self hosted models and/or self hosted APIs, maybe you don’t need MCP to provide a gateway to a secure resource.

If neither of those things are true, you need an authenticating gateway/proxy or a target API that supports single use credentials (and get the model to generate a call to use them).

We can argue whether MCP is a good authenticating middle layer, but not whether one is required.

I always have a Mandela-effect moment about the Dropbox and iPod dismissive comments with Hackernews and Slashdot respectively.

God I’ old

I asked the original commenter to confirm my understanding of what they were saying.

I find it a fascinating alternative view to what is largely well understood (your counter-point).

The thing that stood out is the comment that it’s an approximation of the original data generator (humanity). Early approximations were poor (GPT 2-3, to an extent GPT-4).

I’m not so sure I can reject the hypothesis that such an approximation can be found.

> A lot of people seem to think of the training process as curve-fitting data (the "stochastic parrot" model), but I think of it more as "solving an inverse problem to approximate the unknown source that generated the training data"

I just wanted to confirm your underlying point here: training a model isn’t about finding a function that fits the observed data (even though that’s the outcome) but instead finding an approximation of the unknown source that generated the source data in the first place.

In the case of LLMs an approximation not of a human but of the sum of humanity that produced the training copora?

You may be discounting the tokens generated in the thinking trace but not included in the output to users.

The “problem” is exactly as you’ve framed it: individual 1-1 tutoring unambiguously improves education outcomes by a significant (2 StDev’s in Blooms study) vs any other understood method but delivering this universally is (was?) infeasible.

Bloom looked at other methods used in concert to achieve similar improvements to “solve” the problem that could be delivered at scale.

LLMs may provide a new path to the 2-Sigma improvements without the same delivery problems.