alextheparrot

You also might see different immune preparations, even. Immune cells in different locations make different types of antibodies and which type of antibody recognizes the antigen in the vaccine partly determines which type of resistant you build up.

IgG and IgM are common in the blood, while IgE is found in places like the intestines.

Really sorry to hear about what happened to you

Would recommend to everyone / have bought as a gift a pair of ear plugs that fit onto your key chain. They are like $10 and are honestly the most useful thing I carry on that ring aside from the actual keys. Plus they can be useful in random situations (Needed them to sleep while camping recently because the wind was so strong)

Somehow once I attended to your tokens “stochastic parrots” the rest of your words became much more predictable

Why are we not OK with the program producing the undecidability result?

This is talking specifically about the frameworks people use forwarding telemetry to the developers of those frameworks

It isn’t particularly clear what technical requirements you were working against, but let me give it a shot:

A lot of groups start using Kafka because they have high-throughput event streams they want to aggregate over, and then you just use Kafka for everything because managing a 5 TPS topic alongside a 100k TPS topic is trivial.

In terms of why Kafka is a good fit for that workflow — database writes are unnecessary and oddly structured for raw events we want to expire in a few days, and dealing with buffering blob file writes can cause data loss, so Kafka can really simplify the producer architecture where the producer is also a consumer from a producer who wants an ack. Combine this with how trivial it is to have multiple readers on a pub sub system, and it is easy to scale from 1 to N consumers of a data source without duplicating the data everywhere. E.g. you could have three aggregation jobs that use the same data, one job that writes the topic’s data to blob-style storage for batch use-cases, and a low latency inference job all running from the same data stream.

More or less, Kafka just simplifies scale-out in some cases, maybe not your case, though. If you’re kicking off workers to do ingestion, you might have a system where you are pulling files down at some infrequent cadence (Let’s say every 10 minutes) —- in that case Kafka is likely going to be overkill and feel like a lot of work for an API call that then becomes a tasked tracked in some database.

Aside, but 1,000 IQ is a funny number to me. I’m pretty sure that would be 60 standard deviations from a median human, which has a probability density on the order of 10^-700. Would be hard to come up with enough questions to give reasonable confidence in that distribution — maybe a task for a future intelligence

I had an ex who was deeply into roughly this idea, but for beauty products. There’re many micro-domains where people spend a lot of time trying to figure out where things come from, what their environmental impact is, the production process, etc.

Maybe not a billion dollar business, but hey it is an honest thing to try to sell.

My first recollection of this sentiment pertained to the New York Times, for what it’s worth.

The path to the contrarian is more interesting, who knew

What a crazily vulnerable post, cheers man and hope it gets figured out

This comment confuses me a bit, as AI improving via feedback on its generations introduces no new concerns. Which new intellectual property concerns are introduced via reinforcement learning?

An aside, but people use and pay for Copilot, it is out of the research lab phase.

I think some of your points are myopic — the same thing was said for language models, but RLHF and RLAF [0] clearly show a trajectory of the models getting better at doing this self-improvement process with less “humans are special”ness.

A tool like Copilot can more or less automatically improve via telemetry, I’d expect the same thing as image models catch-up. I’d also expect the human signals to get further and further downstream of the actual creation process (e.g. Gameplay tester reports visual bug versus artist manually edits character)

[0] https://www.anthropic.com/constitutional.pdf

Copilot also makes it easier to use open source framework code.

Being able to ask a model for help writing Spark, SQLAlchemy, Tokio, whatever, actually increases the usability of free code vs proprietary code.

Code is the base model in more recent iterations [0]

[0] https://beta.openai.com/docs/model-index-for-researchers

I don’t think a self-driving hearse will calm their nerves

I don’t understand your point — I’ve been rehearsing that response to the normal and all-consuming test question “Are you human?” since Wednesday.

The core caveat being the yield of “improved … prediction” in the future tense, I assume?

Given that improved structures has sped up drug discovery, I can see where the mistake is made (X has improved Y, therefore X will improve Y)

There is a great deal of GAN literature on the order of 5 years old…

“Trending on Artstation” is a genre curated just like Bauhaus or Romanticism. Just the wider populace votes instead of old noble families and wealthy bankers.

I don’t see the moral significance of the former genres versus the latter

If I build and give away bird houses, that could still be a hobby. I don’t follow how lack of ownership means it can’t be a hobby.

Same issue, Version 15.1 (17612.2.9.1.20)

    TypeError: N.at is not a function. (In 'N.at(-1)', 'N.at' is undefined)

The main idea is that instead of a session cookie you are usually exchanging signed information in this scheme. Your app server or an intermediate proxy validates that the signing was authentic, and then receives some information (Often called “claims”) — such as “This account is an admin role user”, or “This account has user id 1234”.

With those pieces of information, you can do all your application lookups securely. With a bit of middleware to handle the incoming signed data (Often JWT), it probably looks about the same as your session-based authentication. The only difference is that you’ve split out and centralized your authentication (And possibly authorization via claims)!

Is this just an elaborate way to trigger someone else’s AWS code using RBAC instead of an API token?

Seems like the “normal” way I’d go about this would be someone sets up a server, I ping it with token auth that’s been provided, thing happens. Now instead of doing auth through headers we require both parties to be on AWS to use their auth magic?

Guess it could be useful if both organizations are within the same company, so you can just enable app-to-app auth. No one is probably worried about cloud vendor compatibility in that case. No diss on it, but it seems kinda niche?

The book is really quite good. The Checklist Manifesto might also be of interest.

The parts of strategy (for some definition of strategy) that are missing from tactics are (1) Diagnosing problems (2) Setting policy. Things like “How do we solve problems” or “How do we find problems” vs “What is the solution to this problem”.

Maybe something more pithy like: “Tactics are what you use to fight on the battlefield, strategy is what gets you there”?

I feel like this post concludes by validating the counterpoints about what it says isn't strategy. Each of these things show self reflection. If self-reflection is what makes for good strategy, then these are certainly trying to take steps in the right direction.

A vision statement is the product of reflecting on who are we. Checklists and templates are products of reflecting on how we do things. Goals are a product of reflecting on who we think we could be tomorrow.

If I were to venture out, a lot of these things are part of a good strategy. And yes, if you set bad goals or have bad checklists or have a bad vision statement, it might just be that the self-reflection was poor. Imagine if Airbus made a checklist of “[X] Don’t let the plane crash into the ground”, probably not rooted in deep reflection.

The book they reference is quite good, though.

I agree with your position that people will notice and that some attributions will be affected.

It seems like we’ve still given cake to the glutton, though, just without a cherry on top.

I’m not sure how true this is, many types of data have diminishing returns after a few months. I’d be surprised if they lost money compared to not using these methods —- they’ve just lost the tail of incremental value.

Apple doesn’t force me to hide my email, it is my choice to click hide vs not. Apple is just enabling the consumer choice — a choice that a lot of consumers do take (As you noted).

There are good parts about the way Apple operates, they execute well and in a consumer-friendly manner — that’s honestly pretty rare and one reason to like Apple.

QR codes even have error correction so the developer can tune error rate…