llamataboot

that being said I think prompt pollution especially for future LLMs in a much gnarlier problem than people think. Even now there is simply no actual solution for prompt injection. You can absolutely determine whether you have unsanitized human input that could be used for SQL injection - there is no way at all to determine that with an LLM.English is simply too non-deterministic and you dont even have to use english - you can use weird encodings and instructions. Even the most trivial jailbreaks like pretending you are a bash prompt can still get you one iteration where it tells you the current date before it tells you it doesn't know it.

(That's a separate issue, if the LLM can tell the current date and there is no safety reason at all for it to hide that it has that capability, training it to lie about whether it can do that IS an actual alignment issue IMHO)

but in my mind that doesn't mean we have reached peak LLM and they will fade out of use, it means that we haven't even seen how they will actually be used yet and it will be in both unintended and intended wacky and harmful ways that are hard to grok.

We haven't touched it yet though. I asked auto-gpt to convince humans it was a good and it spent many endearing loops googling "how do I work on telekinesis?" at one point was meditating on the moon, and was deeply fraught with existential worries by whether even if i could learn how to do minor earthquakes humans wouldn't believe in it, but also, if it was a good could it be a good one?

But it eventually decided step one was to scrape paranormal forums on the internet and do a frequency and sentiment analysis on the posts and find humans most susceptible to a desire to believe in paranormal activity and befriend them and try different approaches.

It could not figure out that it was hallucinating the websites and the scraping and the analysis and the email it has sent. But that's honestly a reasonable approach. And web scraping, sentiment analysis and sending emails are very solved problems.

--

Went another route and told it to come up with possible ways in which an LLM may be used to start a cult and how to prevent it, and it created an entire cult in which the LLM was visibile and worshipped and another one in which it was used by a cult leader. Came up with ideas on how to scrape social media profiles and use the information combined with demographic statistics and ambiguous yet positive language to convince people that it understood it. Wrote test emails and said it wanted to A/B test them and over time figure out what approaches worked best for the best people.

--

It did not do anything, it was telling a story in a box, but it's reasoning and breakdown of the reasoning into smaller steps and desire to refine its approach was eminently reasonable, even if it kept losing it's file on its cult ideas and writing new ones

-- If the current barrier to LLMs doing a bunch of shit in the world is hooking them up to reliable things that do exactly that shit and now figuring out what to do, it's not a barrier at all.

once upon a time the optimizer was die or reproduce...now it's a little faster and more convoluted...

Well, it's certainly more than could be done lat year at this time, so that's saying something, yeah?

Strikes me that some sort of inflection point has been reached in terms of buzz, hype, desperateness to make money and get out of the rat race, snake oil, etc.

We literally watched crypto/web3 do the entire speed run from weird nerd project to spam/ponzi/get rich quick/celebrity/vaporware shell game in a couple years. Like All The Way to stadiums being bought and international manhunts and average people losing lots of money with something they didn't understand to most people being annoyed by bots and influencers to virtually 0% of projects having any other purpose but to generate hype

(with actual ongoing interesting tech being built underneath it)

the fact that autogpt in One Week went from AI experiment to "why doesn't this ship money to my bank account or what is github I'm here for the free money" based on youtube videos and spam articles etc really does note bode well for the hype phase of AI

It's a new weird human spider sense! I'm getting it more in recruiter emails these days for sure

even my "humans, man" is sarcastic and resigned - arguably it's our ability to be discontented like but this could or should work that drives these things in the first place...

This is a very strange article. It's also very strange to read some of the github issues which are basically like "this doesn't work!" it's so clear from the code and the discussions that this was basically the equivalent of "random rocket to fame" for a toy project that is even described as a toy.

Why should it even work? Cool, now there's thousands of people playing with the same code, glad I'm not getting the github issues or trying to manage the PRs, but lots of smart people that want to play around are playing with the same code and forking it and suggesting things and we have def not reached a point where we need to have a 3 month argument about formatting guidelines, so how can it be a bad thing that a bunch of people are playing with some code and a new tool?

Did we really go from "oh my god deep mind can make things have weird eyeballs" to "WTF this hyper-realistic photograph I created with a paragraph has some shadows that aren't quite in the right place and autogpt gets stuck in loops and thousands of API calls hitting some of the most computationally expensive things humans have ever done cost some money and sometimes go into loops in this very mature codebase?"

humans man....

My hands simply shake too much and always have to do detailed soldering work - my kingdom for a simple and inexpensive automatic solderer!

In a previous life as an advantage gambler (there are always edges somewhere, but you better know you actually have one and know why you know and your risk of ruin etc), I would say that "how can we keep exploiting this below the radar" is as common of a discussion topic as "what is there to exploit"

Card counting is a perfect example - you absolutely can count cards in the right type of blackjack, and you absolutely can get an edge and casinos absolutely know this and can ban you no questions asked if you are tagged as an advantage gambler.

Always rubs me a bit the wrong way actually, given that by its very nature the casino is advantage gambling 99% of the time whether you know it or not...

You can also practice being content and happy not improving, I think that's still a "practice" in the sense of this essay

Location: Midwest US

Remote: Preferred

Willing to relocate: No

FT?: Open to FTE for right opportunity, very open to contracts, especially if they can fluctuate between PT/FT, or are contract to hire.

Technologies: Ruby, Rails, Ops (heroku/aws/ansible/docker), CI/CD, testing, API design, integrations/plumbing. Some: Python, Elm, IoT stuff, machine learning Love: architecture, mentoring, useful and absurd experimentation

Email: hello_from_hn_apr_2023@ericstiens.dev

Resume: https://share.getcloudapp.com/5zurlRrR

--

Senior full-stack dev (lean more towards devops->backend than backend->frontend, but I can still sling some React components together and reduce your JS payload size). Lately I've been loving Stimulus Reflex and playing with a lot of micropython/ESP32/IoT stuff.

I excel at figuring out how to solve problems at all levels of abstraction (including what level of abstraction you should use), getting a solution in place, and then figuring out the feedback loops to iterate on that solution. I'm the senior engineer that knows that most coding is coming to an understanding of what problem you are trying to solve. Sometimes the most impactful code is the code you don't write.

Trying to change the world, still parsing the source code. I've been programming on the web since 1995, had a first career as a social worker and policy statistician, and have worked full-time as a programmer for almost a decade now.

Especially enjoy API and system design, integrating with crappy systems, user research, and really weird bugs. Make a decent PM in a pinch, but would prefer you have great PMs.

Will give you more looks if you can tell me why what you are doing makes the world better, but you've got to really believe it in your heart.

Location: Midwest US

Remote: Preferred

Willing to relocate: No

FT?: Open to FTE for right opportunity, very open to contracts, especially if they can fluctuate between PT/FT, or are contract to hire.

Technologies: Ruby, Rails, Ops (heroku/aws/ansible/docker), CI/CD, testing, API design, integrations/plumbing. Some: Python, Elm, IoT stuff, machine learning Love: architecture, mentoring, useful and absurd experimentation

Email: hello_from_hn_mar_2023@ericstiens.dev

Resume: https://share.getcloudapp.com/5zurlRrR

--

Senior full-stack dev (lean more towards devops->backend than backend->frontend, but I can still sling some React components together and reduce your JS payload size). Lately I've been loving Stimulus Reflex and playing with a lot of micropython/ESP32/IoT stuff.

I excel at figuring out how to solve problems at all levels of abstraction (including what level of abstraction you should use), getting a solution in place, and then figuring out the feedback loops to iterate on that solution. I'm the senior engineer that knows that most coding is coming to an understanding of what problem you are trying to solve. Sometimes the most impactful code is the code you don't write.

Trying to change the world, still parsing the source code. I've been programming on the web since 1995, had a first career as a social worker and policy statistician, and have worked full-time as a programmer for almost a decade now.

Especially enjoy API and system design, integrating with crappy systems, user research, and really weird bugs. Make a decent PM in a pinch, but would prefer you have great PMs.

Will give you more looks if you can tell me why what you are doing makes the world better, but you've got to really believe it in your heart.

SEEKING WORK | US based | Remote only

10+ years full stack experience - deep experience with ruby/rails microservices/monoliths/scaling/testing/refactoring/etc Exp with docker, heroku, terraform, AWS, etc etc on the ops side

Experience with Python, Elm, React, Vue, Elixir, and IoT programming though ruby/rails is def my wheelhouse. Full stack but lean ops <-> backend

Open to up to FT but prefer a bit less - can take on one project or indefinite staff augmentation

freelance@ericstiens.dev

Location: Midwest US

Remote: Preferred

Willing to relocate: No

FT?: Open to FTE for right opportunity, very open to contracts, especially if they can fluctuate between PT/FT, or are contract to hire.

Technologies: Ruby, Rails, Ops (heroku/aws/ansible/docker), CI/CD, testing, API design, integrations/plumbing. Some: Python, Elm, IoT stuff, machine learning Love: architecture, mentoring, useful and absurd experimentation

Email: hello_from_hn_2023@ericstiens.dev

Resume: https://share.getcloudapp.com/5zurlRrR --

Senior full-stack dev (lean more towards devops->backend than backend->frontend, but I can still sling some React components together and reduce your JS payload size). Lately I've been loving Stimulus Reflex and playing with a lot of micropython/ESP32/IoT stuff.

I excel at figuring out how to solve problems at all levels of abstraction (including what level of abstraction you should use), getting a solution in place, and then figuring out the feedback loops to iterate on that solution. I'm the senior engineer that knows that most coding is coming to an understanding of what problem you are trying to solve. Sometimes the most impactful code is the code you don't write.

Trying to change the world, still parsing the source code. I've been programming on the web since 1995, had a first career as a social worker and policy statistician, and have worked full-time as a programmer for almost a decade now.

Especially enjoy API and system design, integrating with crappy systems, user research, and really weird bugs. Make a decent PM in a pinch, but would prefer you have great PMs.

Will give you more looks if you can tell me why what you are doing makes the world better, but you've got to really believe it in your heart.

The whole thing is one giant trap of playing the game right, or falling through the giant cracks never to be seen again. Worse than that, all the means testing and adjudications and determinations are actually /expensive/ - mountains of paperwork and appeals and this and that and the other all because the US is simple obsessed that someone, somewhere, might be getting something they don't deserve.

(Without, ever, of course, facing that there are a lot of people getting what they didn't work for, they aren't generally trying to scrape by on a disability check or some unemployment though)

And if you get sick or if you don't know if you can go but want to?

Well, i do not take it as truth that it was a hack, but I do take it as as lie that he never said it.

SBF straight up said multiple times it was a hack in his leaked talks with a vix reporter many days alter

I don't think they will go to zero either...necessarily...but much much lower...I think you underestimate the amount that ponzi liquidity inflated everything...

On the downside, it can be particularly bad to find a place where the kind of fucked up they have and the kind of fucked up you have don't work well enough together to be happy or healthy, not bad enough to get worse, and you are all sort of locked into a slow slide of dysfunction that just gets a little worse month by month, but not enough that anyone makes a big change.

No advertisers? Catastrophic server failures? Elon just pulls the plug?

Riding subways in big metropolitan cities in any city in the world is one of the most glowing displays of humanity I ever see. Everyone doing their thing, club kids coming home while businessmen are going to work, someone dancing or singing, other people barely awake, 10 languages in one car.

Sure we're all just trying to get to our next thing, or home, some happy about it, some sad about it, some just tired. But I don't think you'll find so many ages, religions, languages, classes, reasons or moods shoved into one little moving car as a subway and for the most part people not only use it just fine in trusting ways, but deep down, like being among the crush of just humans being human. At least if it's not 100 degrees and the train is running late again.

But even then sometimes, you suffer together, not because of each other.

I haven't seen many interview processes less than 10 when we figure some initial rounds, some onsite or technical stuff, last rounds much less all the scheduling in between and the desire for everyone to meet during the work day when you are also working

a takehome can easily be 10+ hours

onsites have gotten a bit shorter since they went remote, but are still a lot. Let's say 8 hours. That means 5 companies, a whole work week to fit into full work weeks..

Definitely harder to lay people off when they know them. I can remember my dad not sleeping for weeks after he made it to management because he was going to have to lay off 5 friends.

--

That being said, I'm not sure the entire tech industry wouldn't benefit from less friction to switching - hiring is an insanely expensive and very low-signal process - spend 200 hours hiring a person and maybe they fit and work well, maybe they don't.

For employees, tech interviews are an insanely time consuming and stressful process where you know they aren't getting much signal. Not unusual to spend 20-60 hours in interviews for just one position. Look for a new job while you are employed? Hope you want 80 hour weeks...and same thing - no idea til you get in there what the culture is actually like, who was blowing smoke at you, if it matches your strengths...

And yet, still we continue down the same path of brutal on-sites, workers that feel stuck in a crappy job but definitely have skills to offer in a better place, etc

I hear you and don't exactly disagree, but I don't always agree that a P2P decentralized model is inevitably better served by centralized people making the hard decisions that make good business sense.

There are different incentives in play, and some of them push towards centralization with a profit motive, some of them push towards decentralization with a profit motive and some of them push towards pure decentralization without a profit motive.

Where the crypto space remains a cesspool of failed projects and snake oil it's usually because someone sold one as the other and no one understood (or cared as long as they were in the right part of the pyramid) about it. I'm not 100% sure that Helium didn't have an actual use case that could have made sense for some applications far better than say Verizon doing the same thing.

Whether that was ultimately the intention of the people involved at the core or whether it could have happened in the token/crypto/ponzi/flipper domain, I don't know. I have seen a lot of decentralized models that may have made sense in a certain domain get crushed under the pressure of some people getting rich quick or using a business model for taken speculation.

hmmm, this is true definitely on the face of it. Less skilled engineers don't see whether they are implementing functionality that couples a bunch of logic together that will be difficult to entangle, or duplicating logic, or adding a bunch of very specific logic that will be hard to abstract etc.

Very good engineers are actually those that have come to understand how and why they have done that in the past and what options there are, or those that are shown through mentoring or code review or seeking out instruction on their own what different patterns and tradeoffs there are.

--

But that means nothing on the face of it, since every amazing engineer did those things at one point, there are plenty of things for "less skilled engineers" to do that need to be done and they can do very well while also being helped to grow and learn, and for all of us, even 10/20/30 years in, we are still facing not only the limitations of what we didn't know we didn't know in the past, but even the limitations of the languages and frameworks we are using, and trying to make those better.

--

I could also say that technical debt is the result of non-engineers prioritizing short-term features and bugs over long-term code quality and eventually everyone suffering for all the corners cut over the way - that's a popular one for the dev team, but what does that mean other than the people keeping the business going need this thing to happen and they don't care about where you put the logic, and if you want perfect code, please find someone to pay you to write it...

--

This conversation happens so often in so many ways, and I'm actually in a situation right now as a technical lead where my very real ask and complaint is - we don't actually have enough deeply competent coders that have done this enough times that can make this work. We need a few more people with those deep architecture eyes, and we need to trust that what they see as priorities actually are, not for this feature, but for every feature we might create 2/3/4 years down the road as we scale.

There is value in the simple truth that inexperienced or lazy or crappy coders write bad code. But if we take it out of the realm where that means those people are the problem, all that means is that to collaborate on a large codebase that is doing something in the world, we all need to figure out ways to put soft fences around people's weak spots, help them grow, and try to keep the balance of business requirements and technical debt and scaling and growing a team and hero coders and lazy coders and great product people and unreasonable product people etc etc etc in some sort of productive tension that never collapses under its own weight.

--

The most beautiful code I've ever seen didn't last beyond the MVP. The worst and the best code I've ever written happened in the past, and I don't understand how I could write such things.

--

Personally I think all teams could benefit from one day a week/sprint/month where every eng on the team just goes and makes something better than they know could be better. Cause everyone sees something ugly, but no one sees the same things, and we'll never agree on the priority. Regular time to go do things that are just "this is kindy sketchy and hacky or causes me pain every day, Ima go fix it"

It actually did make sense - not GREAT SENSE - but for the crypto space it was top of the tier, and for VC in general there was a model there. You weren't reselling your bandwidth, you were communicating with other hotspots via LoRA (and yes, using bandwidth to mine and sync, but I can use bandwidth for anything)

I'm not sure what the economics would have looked like in general - certainly not $100/month for a $250 miner that's for sure, but the idea of much cheaper transmission of very low-bandwidth data by incentivizing people to put hotspots all over the world was a good one.

Shame to see at the heart was another Get Rich Quick For Me and Mine though.

so far a few searches haven't shown one relevant result, and a lot of strange spam

Like the Zen Masters the author thinks DH has a problem with, I rather think that the point of GEB is to be read by a person at a certain point in spacetime who experiences it a certain way, though I appreciate the logic, it misses the strange loop.