cryptoz

Starship without Booster has had many successful suborbital tests, but with Booster only 1 attempt. Much of that attempt was successful but it did have critical issues and they had to remotely blow it up, which also initially failed.

If this gets to orbit or close enough, that will be considered the first successful real launch.

I actually really like this definition and will be giving it some thought. But right off the bat, that’s not how most people will see it - and so while this definition is certainly thought-provoking and useful, it doesn’t specify much that’s relatable to other tasks and therefore I think will always be a niche definition.

An AI that can make a better version of itself may not be able to communicate in any human language for example; and that is now a de facto requirement for most people to see something as AI I think.

Something to note here is that copilot is pretty old tech at this point already - GPT4 API is just absolutely worlds better, and it is also about to be old news soon too! With GPT5 coming probably next year, we don’t know what advances it will make but surely basing your understanding of it on copilot will leave you shocked soon enough.

Well for a quick point of comparison there are very numerous and very strict rules about what you’re allowed to do with a Mac. Or iPhone. Or Android. Or windows device. All of them forbid you from learning about how they work, much less changing how they work. Not sure about legality but certainly it’s against the license. Which is what you transacted right? You don’t “own” your iPhone for any reasonable definition of the word “own”. But you bought it, or so you think, but what you really did is pay money and sign an agreement. I doubt Apple thinks that you actually own the device once you buy it.

I can’t wait for the time when Starship is proven and launching frequently - though there is a lot to be said about the excitement of right now, when it hasn’t yet had a successful launch.

Really got my fingers crossed for a true rapid drop in cost to launch mass to orbit.

I remember that moment! I was out at a bar or something at the time but I was prepared and had my laptop with me haha. I was mashing the up arrow and enter to make sure I didn’t miss it.

> When passing an entire design specification into an LLM and receiving a new representation token by token, generating a response would take several minutes, making it impractical.

Woe is me, it takes minutes to go from user-designed mockup to real, high-quality code? Unacceptable, I tell you!

But seriously, if there are speed improvements that you can make and are on the multiple-orders-of-magnitude then I do get it, those improvements are game-changing. But also, I think we're racing too quickly with expectations here; where minutes is unacceptable now when it used to take a human days? I mean, minutes is still pretty good! IMO.

OpenAI recently updated their limit documentation so there are tiers now. You automatically scale up through the tiers but you may be able to jump the line a bit by paying some extra. But I think you will have to wait 7-14 days since first payment before going to next tier. Try finding the usage and limits docs in your account.

Edit: Oh I see I'm in Tier 3 and the RPD limit on vision preview is still 100/day. Guess that's just it for now. It is labelled preview after all.

Oh neat, I hadn't heard of them yet. Will follow that for sure. I miss Planetary Resources - I got hyped on that, but I guess they were too early.

The probe only goes 550 AU out in the opposite direction from the planet so we can use gravitational lensing to see up close!

Yeah it’ll take a lot of money for sure. Needs new propulsion like nuclear thermal or something to get to target distance in our lifetimes.

Would be very expensive but I think cheaper per mission, if you start off with a plan to send a lot of them.

Might take in full some tens or hundreds of billions. YC I am aware will not fund on that level haha, but maybe they would have an eye for wanting to start it off.

I've been idly wondering if it's worth it to apply to YC next batch with the idea of launching dozens of 550AU missions for solar-gravitational-lens HD photographs of nearby exoplanets. I've been wondering a lot recently about our visibility to potential life out there; they may well be watching Earth in HD since we can imagine how we might do that too. So I think a great step would be launching a bunch of long missions that will eventually return us HD video of exoplanets within like 200 ly or more distant maybe. That will reveal a lot of info about close-by worlds and may produce copious evidence for life on other planets.

Would be very expensive, take a few decades at least, and the profit comes from...governments? Haha not sure about that yet. Might have to pitch it as a planetary defense company and also build tech to zap asteroids etc.

Basically, NASA is doing the great hard science obviously, but is outdone in pacing and tech by SpaceX and other startups; NASA plans to send 1-5? 550 AU missions eventually. But they're in no rush. I want to rush it.

Yeah, I defintely agree with that haha. I'm new at this and haven't even run my own SD or any open source model on my own hardware yet. I'm sure it's not that hard, but I don't think my CPU-rated Linode is up to the task. Haven't looked into renting a cloud machine that can do it quickly but I'm sure it exists. I absolutely want to move the LLM away from OpenAI GPT and the image generated away from DALL-E. But for right now this was the fastest way I could get the idea online, so APIs it is.

But I'm 100% on board with the open source models and I definitely want to get that going. Yeah the DALL-E 3 images are 4 cents per image! Phewph!

Okay it's 1pm PT. How are you testing when you get the new features? Just running a curl or something until it works? :)

Aha, makes sense, thanks :)

For DALL-E 3, I'm getting "openai.error.InvalidRequestError: The model `dall-e-3` does not exist." is this for everyone right now? Maybe it's gonna be out any minute.

I see the python library has an upgrade available with breaking changes, is there any guide for the changes I'll need to make? And will the DALL-E 3 endpoint require the upgrade? So many questions.

Edit: Oh I see,

> We’ll begin rolling out new features to OpenAI customers starting at 1pm PT today.

Deserve to lose the next space race because of a minor delay? Too much hyberbole.

Honestly, I'm a huge huge huge Mars fan and spaceflight fan and all-things-space fan but I don't care one inch about the environmental review delays.

WTF is the point of going to Mars if we don't even care about the endangered wildlife we're killing on Earth while we go? I mean, yes, some wildlife will die in the human pursuit of spaceflight. It already has, lots of it. Humans too.

But at the end of the day, we ought to care. The whole point of going to Mars is to spread life to other words. At the very least we should observe and care about the ways we kill it here especially when in the pursuit of space travel.

The real trick is actually enforcement. Nobody obeys the rules because none of the rules are enforced. So sure, make the law that prohibits just certain vehicles from doing a right on red. 100% guarantee you that the drivers of those cars will not obey the new rule and 100% guarantee you that cops won't enforce it.

Happened in Seattle today actually, I think. Car running a red light, possibly doing a right-turn-on-red, a bus hits the car and a pedestrian died.

As the years go by I'm getting more militant about enforcement of driving laws. Nobody obeys the rules because there is never enforcement. And it's getting worse and worse, anecdotally. I haven't seen a car actually stop at a stop sign probably in months. It's all straight through, or slowing down. Never stopping. Even if there is someone crossing in front of them, the cars rarely stop. They will even accelerate towards you while you cross, hoping you'll run out of their way so that they can get somewhere 2s faster.

Edit: Another aside, in Montreal I think there's a sign as you enter the city that says "Welcome to Montreal No Right on Red" lol. Some cities have "A beautiful and welcoming place" or whatever but MTL goes with "No right on red".

Some extra context that didn't fit in the description above: Each game turn picks an OpenAI model to use - runs on a different GPT model depending on my rate limits, current usage, etc. Interestingly I haven't found much difference in game quality between GPT-4 and GPT-3.5. IMO the coding quality is dramatically higher in 4 but 3.5 seems more than capable for this game. There's no UI switch for you to choose your model yet but I plan on adding something like that soon.

The site I'm launching this on is a Space / AI experiment site. The main page is DALL-E images generated from prompts that are generated by GPT-4, for example. I am hoping to add more fun new experiments like a constant GPT-4 conversation where multiple agents with different specializations talk to each other about space exploration (ie, "You're an astrophysicist talking to a rocket enginner and a biologist about exoplanet exploration" etc).

About the image generations: Unfortunately DALL-E 3 API access isn't open yet (Monday?!) so the image quality is less than what you'd hope for - but that should change soon. Also, my GPT-created image prompts for the visualizations by DALL-E are not great. I think it's currently only reading the last turn in the game to make the prompt, so there is not much hope at continuity. Working on using one of the larger-context GPT-3.5 models to generate better image prompts (I don't have GPT-4 32k context access yet)

Yeah you can only dream about 1 thing forever. Can’t have multiple things you dream about unfortunately :(

AI is already it.

All this research from Google and still no discussion of using their access to billions of barometers from Android devices? sigh.

Some for sure but all the algorithms (simple ones) and such is also done successfully by ChatGPT.

That is a great point and I will definitely share my prompting experience and some real prompts in a blog post this week. I'll come back here and link to it when ready.

I’ve been playing with ChatGPT code generation to make entire sites with flask, python, html+js+css, backed with SQLite db and it’s amazing. I’ve had it write like 5k lines that are all live in prod and working (not much traffic lol but still).

A huge huge factor is knowing the limitations and getting better at prompting. And identifying likely hallucinations and asking for risks etc.

I’ve found it best with tech I don’t know well (I’m an android dev using it to make websites, something I haven’t done myself in like 15 years).

Most of the coolest stuff for me is help with sysadmin and running the server. The ability to debug gunicorn errors is great.

I do have to modify the code it outputs as the project grows and it loses context, but honestly the context limits are the biggest hurdle for bigger projects and those will be lifted soon.

Edit: Most recent site I made with like 95% code from ChatGPT is https://cosmictrip.space/ which generates prompts with GPT-4 that are then used to generate space images with DALL-E.

It's a simple site but there is a secret adventure game I'm working on (GPT+Dall-E) that is open-ended image+text AI-driven game. I'm hoping to launch before Nov 6 with DALL-E 3 API (hopefully...!). The adventure game is also written like 95%+ by ChatGPT.

I've had such great success with it coding that I'm using the GPT-4 API with an agent I'm making (everyone is huh). I have function calling hooked up to generate structured subtasks that the agent can then write the code for, and support for files to include context, chat with your code, etc. It's not ready to show but the GPT-4 code generation abilities are really incredible - but you have to be experienced at prompting. Your first prompts aren't likely to be great, which is why I'm hoping my agent can have success. The idea of the agent I'm writing is a Jira/kanban style board where you have AI coders assigned to tasks that you can approve and modify etc. The tickets should automatically move across the columns as the AI checks the work etc.

Can’t risk losing the AI war entirely. Much better to be embarrassed to the tune of $2B than to have no profits from future AI success.

Just chiming in to say I was super into Uplink for a bit around 2003 as well. Absolutely loved that game and couldn’t ever really put a finger on why I loved it so much.

Some of the worst wildfires in the US west over the last couple years were set off by idiots with fireworks FWIW.

And maybe a gender reveal one as well? Can’t remember but it seems likely true.

There's that protected-from-open-ocean barge lane (I forget what it's called?) that I think runs right from Brownsville to Cape Canaveral, or close enough, I think they plan on using that.

Edit https://en.wikipedia.org/wiki/Intracoastal_Waterway