SmooL

I've been having a good time with Spec Driven Development, and it directly addresses the issue of needing the understand.

The whole idea is that you specify exactly what you want in some SPEC.md file. You can of course nest them, have multiple, etc, but the core idea is that the SPEC file is the source of truth, and all the code should be able to be generated by a competent agent into the working product you want. The SPEC file(s) should contain all the details and behavior you care about, and anything you don't care about is up to the agent to decide. If you don't like what the agent picked, _put it in the spec file_.

Critically, _you_ must write the SPEC file. You ensure understanding by doing so. You can of course ideate with the agent, but it's your ideas, in your words, specified by you. This also makes it a great source of documentation when you come back later and have to remember wtf is going on in this codebase.

I can trivially just print any AI image I want, then take a "verified" picture of it with my camera. That seems like a pretty large failure point.

The thinking goes: - any job that can be done on a computer is immediately outsourced to AI, since the AI is smarter and cheaper than humans - humanoid robots are built that are cheap to produce, using tech advances that the AI discovered - any job that can be done by a human is immediately outsourced to a robot, since the robot is better/faster/stronger/cheaper than humans

That OpenAI demo is available right now to subscribers; I have access to it

Ironically, it seems the article itself was written by AI

There's no good reason for BTC to be set up to use that much energy. There are viable, proven alternatives like proof of stake. I'd also be more forgiving if BTC only required a given amount of power, but no, the system is designed such the power requirement is related to the cost of flops/watt - it's specifically designed such that the system will use more power as power becomes cheaper.

This is/could be a legitimate use of blockchain, e.g. see https://gwern.net/timestamping

Isn't that a bit of a cop-out though? They're just redefining the word "intelligence" to be something else (here, the ability to learn new things). That's fine and all, but that doesn't answer the question we actually care about, which is the absolute magnitude of the ability. It doesn't matter how "long" an entity took to learn, or how "efficient" they were in learning - at the end of the day, AlphaZero will crush Magnus.

It was a multi year effort mainly around 2015/16 but extended into 2018 IIRC

Genuine question: are the health risks really that bad? Plastic has been around for decades - surely also then these micro/nano plastics have been around for decades? I can't help but feel like if there were obvious and large health effects that we'd have noticed by now.

This is either brilliantly absurd, or ridiculously stupid, depending on your POV

This is not a fair comparison - Tesla is comparing miles driven by their systems, which are inherently limited in scope to "easier" scenarios (only lane assist/only on highways/ect), against total US average, which includes _all_ gnarly road situations.

Yes, but transmission lines can only go so far, and you still lack the ability to arbitrage over time instead of just spatially. E.g. from a solar power POV, it's night everywhere in a given region at the same time.

> My gut feeling is that transmission lines would still be cheaper

Transmission lines are great for moving electricity, but only if there's demand for that electricity _right now_. Otherwise, you have to store it - which is a problem, because battery tech right now isn't great (or rather, it's not good enough for grid-scale requirements) . This iron powder could be thought of as a "battery". It might be harder to move than compared to a transmission line, but it's _stored_ energy and can be redeemed at a later time.

1. It's opinionated, so there's often only one way of doings things. Largely, the "one way" is a good way, so people appreciate the forced consistency 2. It's simple. It is very easy to read and write. It is hard to shoot yourself in the foot. 3. It's powerful. They have a few core abstractions that compose well (generic io, http stuff). 4. It's fast. It runs fast because it's compiled, and it compiles fast because it's simple.

Me personally: I appreciate the simplicity of it. It's a great language for working with in a team. I wish it was more functional, and had better ways to handle errors, but the simplicity of it all was a breath of fresh air using it in a working environment.

I've thought about doing this as well, but I haven't tried it yet. Are there any resources/blogs/information on various strategies on how to best chunk & embed arbitrary text?

Doesn't this completely miss the point? DCA is about reducing volatility, not maximizing return. Your expected value is higher without DCA, but it's not about the expected value - it's about tightening the stddev of possible outcomes. "A bird in the hand is worth two in the bush".

Yeah this isn't using goroutines at all? I don't see how this is a good comparison of goroutines vs coroutines.

It's not _quite_ the same: you can't call async code from a sync context (hence the color issue), but I can always pass a "context.Background()" or such as a context value if I don't already have one.

The proposed solution is to feed relevant data from a database of "ground truth facts" into the query (I'm assuming using the usual method of similarity search leveraging embedding vectors).

This solution... doesn't prohibit hallucinations? As far as I can tell it only makes them less likely. The AI is still totally capable of hallucinating, it's just less likely to hallucinate an answer to _question X_ if the query includes data that has the answer.

I've been thinking that it might be useful if you could actually _remove_ all the stored facts that the LLM has inside of it. I believe that an LLM that didn't natively know a whole bunch of random trivia facts, didn't know basic math, didn't know much about anything _except_ what was put into the initial query would be valuable. The AI can't hallucinate anything if it doesn't know anything to hallucinate.

How you achieve this practically I have no clue. I'm not sure it's even possible to remove the knowledge that 1+1=2 without removing the knowledge of how to write a python script one could execute to figure it out.

This is slightly misleading - what matters isn't the peak G force experienced, but moreso the integral of force over time. In a roller coaster you might experience 5Gs, but only for a moment. Astronauts experience 3G, but for up to 5 minutes continuously

I think it's more like: with the original GPT 3 model, it's easy to summon _any_ ego. With ChatGPT, you can either summon a) the intended Luigi or b) the unintended Waluigi, but trying to get anything else is more difficult. The theory would be that, in removing all the other egos other than Luigi, they've also indirectly promoted Waluigi

As the article states, the idea is that this would be faster than ships, as well as operational in areas without large bodies of water.

Yes, but if you have many fat tailed distributions, increasing the sample size makes it increasingly less likely that it's _the same person at the end each time_

Going to second Carcassonne, and say that one thing I really like about the game is that's almost purely contextless, in that it doesn't require you to hold long-term strategy/ideas/concepts in your head for extended periods of time. If you showed up to the middle of the game, and were told to play the turn for somebody, largely you could. This makes it a great casual game with friends, as a long continuous attention span isn't really required.

Probably left it running? I've definitely left my factory to idle to build up resources while I'm gone, when I'm playing on peaceful mode.

You could also run headless version of the game on a server and just connect to it for the same effect.

Remember - Factorio is just an idle game with extra steps.

It's far greener per ton of cargo, not absolutely.

Because abbreviations are nice when there's a shared context that is perpetually "in memory" - I don't have to think to know what LA stands for. However, that is not the case when debugging. It is almost certainly the case for the person who wrote the code originally, but it is certainly not the case for the next person to come read it afterwards. IMO code is meant to be read, so I try to never use abbreviations (other than e.g. _i_ as a for loop index, which, given it's pervasive use, falls under the category of "shared context").

I've had something like this happen to me, albeit less serious: I was snowboarding, went over a jump poorly, and found myself totally disoriented as I careened through the air. Every single other the I'd found myself flying through the air flashed through my head at unbelievable speed, in an attempt to search for some past experience to leverage wtf I should do. Ultimately I found nothing, made peace with the fact that I was going to slam into the ground, and promptly did so.

My best at a summary: as the world becomes more connected, more online, and more seemingly absurd, it takes on a surreal feeling that people like to dive into. Fundamentally, people like to tell Stories, and to be a part of them, and to build them up. Some people would probably call that, given a community of participants, a communities "culture". As remote work takes off, as people increasingly interact with automated systems, as the amount of passive social interaction decreases, people turn more and more to these cultures.

Viewing this through a managerial lens gives us "lorecraft", but I'd probably call it something more akin to "organically created culture".

> Cope or Capability? I would argue that it probably started as a cope, and upon people realizing that "this is a thing that's happening", started to treat it more and morph it into a capability.