valine

47 ms·

511.
by valine 4y ago
It’s not Tesla’s fault that CBS runs a story every time someone crashes their Model Y. If you’re reading about a Tesla crash odds are it wasn’t on autopilot. The news is happy to run the story either way because it generates clicks.
512.
by valine 4y ago
I get the feeling that "apocalypse" is code for "shareholders lose value". You can't sell insulin if the AGI finds a cure for diabetes. We know for a fact that there are problems AGI could help us solve today. The d
513.
by valine 4y ago
There’s no need for daylight savings time on the moon. The day night cycle on the moon takes a month, and plus there’s no farmers up there yet.
514.
by valine 4y ago
I don’t believe this was ever in question. You can think of the model as a giant function that takes a list vectors as input and spits out a new vector. If you want the model to remember something, you have to include it in the list of inpu
515.
by valine 4y ago
“Life expectancy is dropping in the US” is a great clickbait headline. Life expectancy has been rising steadily for decades, with a tiny blip downwards the last two years. You think maybe that big pandemic that killed a million people in th
516.
by valine 4y ago
It’s hard to know what’s real and what’s hysteria when the field is so fast moving. Mass hysteria is well documented and experts aren’t necessarily immune. https://en.m.wikipedia.org/wiki/List_of_mass_hysteria_cases I’
517.
by valine 4y ago
I thought it might be about OTC drug purchases. I imagine walmart has a unique window into the ailments of Americans.
518.
by valine 4y ago
Small animal brains can’t write any code at all, but they’re great a moving around the real world. It’s almost like code and locomotion are two separate problems, and failure in one doesn’t say anything about the difficulty of the other.
519.
by valine 4y ago
It’s not just the cost of training the model, it’s the cost of doing inference at scale. ChatGPT boarder line too expensive to operate already. It’s hard to imagine a larger model that both economical and used by millions of people with our
520.
by valine 4y ago
GPT3 is far more accurate than GPT2. Seems reasonable that larger models trained on more data will continue to improve accuracy. I'd also expect larger models to be better at summarizing text, ie potentially fixing the Bing issues wher
521.
by valine 4y ago
In their current iteration the models are very neutered. It’s been demonstrated that GPT models are fairly good at choosing when to perform a task. Obviously lots of APIs and machinery is needed to actually perform tasks, but the heavy lift
522.
by valine 4y ago
It’s not a stretch to refine the model to store summaries in a database I don’t think. Microsoft is already doing something similar where Sydney generates search queries. Seems reasonable the model could be trained to insert $(store)”summar
523.
by valine 4y ago
I didn’t say it works in those scenarios, I said it lets you enable it. It enables in snow and rain yes. The only time it refuses to engage is in extreme weather that obstructs the cameras. It gets confused a lot in heavy traffic but it wil
524.
by valine 4y ago
> Teslas autopilot can only be enabled in comparatively safe and simple circumstances This is incorrect. FSD can be enabled everywhere from dirt road to parking lots, even highways and dense urban environments like NYC. There is no geofe
525.
by valine 4y ago
It’s not far from the output of gpt3. I wouldn’t put it past microsoft to cheap out on the reenforcement learning.
526.
by valine 4y ago
It’s predicting the next word given the text so far. The entire chat history is fed as an input for predicting the next word.
527.
by valine 4y ago
Except it’s already been shown LLMs can do exactly that. You can prime the model to insert something like ${API CALL HERE} into its output. Then it’s just a matter of finding that string and calling the api. Toolformer does something really
528.
by valine 4y ago
How it should work is the model should be pre-trained to interact with the bing backend and make targeted search queries as it sees fit. I wouldn’t put it past Microsoft to do something stupid like ground gpt3.5 with the top three bing resu
529.
by valine 4y ago
A tree can only play 20 questions if it’s asking the questions. ChatGPT can go the other way.
530.
by valine 4y ago
A parrot can’t play 20 questions, emulate theory of mind, or implement merge sort with pirate themed variable names.
531.
by valine 4y ago
There is evidence to suggest that LLMs have recently passed a threshold where their performance can be improved by training on their own output. https://arxiv.org/abs/2210.11610 I expect your concerns are likely unwarr
532.
by valine 4y ago
Wealth inequality is the biggest risk IMO. It might come down to whether or not we get a competitive open source LLM. Ironically, AI safety is being used as a cover for not open sourcing these models.
533.
by valine 4y ago
The thing is LLMs can be tested and inaccuracies corrected. In a year when the ChatGPTs of the world have API access to augment their already impressive capabilities, I doubt there will be many complaints about accuracy. You can’t run bench
534.
by valine 4y ago
Since we are talking about an imaginary technology here, why not imagine the robots don’t experience consciousness? Problem solved.
535.
by valine 4y ago
A multimodal LLM using tools definitely feels like the closest we’ve ever been to AGI.
536.
by valine 4y ago
So everyone who wants the new port should be forced to buy a new case? Have we really gotten to the point where we ridicule people for not participating in throw away culture? People should have the tools to update and maintain their stuff.
537.
by valine 4y ago
ChatGPT disagrees that it has theory of mind. “As an AI language model, I do not have consciousness, emotions, or mental states, so I cannot have a theory of mind in the same way that a human can. My ability to predict your friend Sam'
538.
by valine 4y ago
I would apply that logic to anyone responsible for honest mistakes. Mistakes are a valuable opportunity for learning, and learning doesn’t thrive in a culture of fear. The same system should definitely exist in law enforcement, to the exten
539.
by valine 4y ago
If the goal is to improve safety firing people is almost always the wrong response. That juvenile preoccupation with punishment only serves to create a culture where mistakes are hidden rather than understood.
540.
by valine 4y ago
Something like searchQuery+=“ reddit” should do the trick.

More ›