To coordinate that many people, you would need either:
a political party that would coordinate global actions via some sort of hierarchy
a disaster of some kind that affected enough of the population that the entire world could be convinced to act all at once, or at least in quick succession, but still didn’t take out all of our communication structures so decentralized communication would still be possible.
We saw how Covid worked out so I think the likelihood of everyone not only acting at once, but also in unison, because of a disaster is quite small without a party to coordinate. There need to be constraints on behavior with levers of power to pull and enforce those constraints in order to get literally billions of people to do the same thing at the same time. I don’t see a way around it.
Could you point me to some reading on uses of AI for disabled people? I’m familiar with a variety of aids and the general idea of “if it looks useless, it’s not for you to use”. I’m just not familiar with what the use cases are.
Unrelated to the question of ableism, this specifically logic that’s pushed by tech companies in general, that their decisions were inevitable and therefore there is no point in questioning them.
Look at how modern LLMs work. They’re trained in large data centers owned by private companies using giant corpuses of data that were largely obtained without the permission or knowledge of the people who created it. Then, to use them, the weights are loaded into an amount of memory that’s out of reach for most consumer desktops and users must call into the LLM using an API. Working memory of a conversation doesn’t persist in between messages or tool calls, so the entire history must be loaded into its context window on every call. In other words, all the “learning” for these models must take place up front in training and outside of taking context into account, it doesn’t actually adjust to learn new things about the world. There are workarounds for this, of course, to simulate the experience of interacting with something that can learn, but they have their limitations and aren’t reliable yet. I could go on. Running probabilistic process on deterministic hardware is an area that we may see more work on soon.
Every single step of that description had alternatives that would be more likely to be chosen outside of a capitalist system. They could be more eco friendly. They could be more efficient. They could be more powerful and learn from your interactions in way that persists. And a lot of these changes would delay the exposure of LLMs to the general public and see them spending longer in academia. But that would be okay because we wouldn’t have the profit motive at the center of this inflating a giant bubble that’s poised to pop and flatten the economy. Bottom line is this stuff was pushed out and hyped up well before it was ready and well before it was able to be scaled up ethically and with the working class in mind. None of this was inevitable.
I just recently got a bunch of medical stuff and psychiatric stuff sorted out after decades of barely being able to function. So I’m finally feeling “normal” and the state of our politics has me stressed to the point of action, but not to the point of curling up in a ball and doing nothing, which is unusual for me. Very surreal to be functional in a nonfunctional world.
Hectic and our teachers are constantly complaining about how kids can’t do basic reading comprehension tasks. Like copy a sentence from the board into their notebook and then repeat it back to you in their own words. There are 13 year olds who cannot do this. There are adults who can’t do this. The majority of Americans read at or below the level of an 11 year old.
I wrote an app for my wife and it was really sad watching her just fumble past bugs instead of pointing them out when I was literally watching over her shoulder to get feedback on what needed fixed. I had to tell her several times, “No, don’t just keep reloading. What’s wrong?” Like we’ve all been trained so hard to accept shitty software that even when I could fix stuff easily I know people are just passively accepting the bugs.
The SOTA changes every couple weeks, but Claude’s been very dominant for a while, yeah. There’s currently a lot of hype around GPT-5.4, but even then there’s a caveat that Claude is still better at UI.
I just personally find Cursor to be pretty buggy. But I think the Replit mention is more of a tell that someone vibe codes but doesn’t actually code. It’s been advertised to people as a way to build end to end apps without any coding experience. And to be fair, they’ve done a good job of building on the past decade of work in the Typescript community to make an entire app end to end type safe and therefore checkable by the compiler. Convex has done something similar in a way that I prefer and in my experience LLMs are very good at working in Convex projects as well.
Really at the end of the day I was just being pithy. Kind of poking fun at how much of a moving target SOTA is.
Yeah I thought she didn’t endorse Bernie because he was so sexist. Maybe she hasn’t seen Platner’s Reddit