You're not wrong, but the need to please the user is still paramount, otherwise they'll just do something else. This is why TikTok is eating everyone's lunch.
I don't agree with this and to answer the question you originally asked me, I do think users are consuming things they don't actually enjoy. The goal isn't to please the user, the goal is to not bore the user. If you talk to people I'm sure you'll find a lot of the music they listened to isn't "enjoyed" so much as it is inoffensive background noise.
It's not surprising that some people are mindless consumers, but it's not useful to assume the majority is, especially of paying customers, and competition exists.
You're assuming it's not useful because it doesn't bode well for your argument. What makes you think assuming the majority aren't mindless consumers is useful?
The hypocrisy lies in the fact that the philosophy of Ayn Rand - that an elite few held up society and the rest were pretty much just parasites - has been used at great length to justify the gutting of social programs.
Please read my comment in good faith.
There is no contradiction with Rand’s philosophy here.
According to her framework, the state stole from her throughout her life. Using public assistance is merely retrieving a small piece of that stolen money.
I agree that it was in her philosophical framework to accept social security - apologies if my comment seemed in bad faith due to that not being clearer. The irony does not lie with her, but rather those that use her philosophy to eliminate the safety net that she herself ended up using.
Sure, she could have used the money she had put into social security to invest, and maybe would have come out better off. But for those of us who see how public services can enrich an entire society, there is irony to how this all played out.
"The irony is with those who believe that thievery is wrong. She obviously didn't believe what she wrote because her actions reveal she believed in stealing your stolen property back from a thief, which is itself thievery."
"The irony is with those who believe that thievery is wrong. She obviously didn't believe what she wrote because her actions reveal she was OK accepting when the thieve gave her your property to make up for the theft she suffered earlier"
FTFY.
She didn't steal from the thieve, she became complicit with the thieve stealing other people's work to get their money back (gracefully handed by the thieve).
She believed that even wealthy kids that just live off their trust funds were parasites too. It was about consuming vs producing, not elite vs non-elite.
I think of the foundational model like CPUs. They're the core of powerful, general-purpose computers, and will likely remain popular and common for most computing solutions. But we also have GPUs, microcontrollers, FPGAs, etc. that don't just act as the core of a wide variety of solutions, but are also paired alongside CPUs for specific use cases that need specialization.
Foundational models are not great for many specific tasks. Assuming that one architecture will eventually work for everything is like saying that x86/amd64/ARM will be all we ever need for processors.
Not to be too pedantic, but code is a kind of specification. I think making the blanket statement "Prompt is code" is inaccurate but there does exist a methodology of writing prompts as if they are specifications that can reliably converted to computational actions, and I believe we're heading toward that.
I’m all for this movement provided it’s actually focusing on the rights of individuals rather than empowering corporations to own and operate massive amounts of computing power unchecked. When I first read the article, I frankly assumed this was meant to limit regulation on AI. From what I’ve read in the law that doesn’t seem to explicitly be the case, but given the organizations involved, I fully expect to see more in that vein.
But this isn’t a suggestion to turn away from AI threats - it’s a matter of prioritization. There are more imminent threats that we know can turn apocalyptic that swaths of people in power are completely ignoring and instead fretting over AI.
Really appreciate the detailed article! I was on the team that shipped D3D11 and helped with the start of D3D12. I went off to other things and lost touch - the API has come a long way! So many of these features were just little whispers of ideas exchanged in the hallways. So cool to see them come to life!
That requires the operating system to “hint” to the display that there’s no refresh necessary and for the display to shut down during those times. That’s currently not supported as these kits just take a video signal, but it’s something being worked on for a future version!
Edit: You work on that stuff, right? Then this armchair experting feels silly, just imagine it's for other readers.
It seems much more practical (if a little less power-efficient) to implement the no diff -> no refresh logic for screen regions in the display hardware. The RAM and logic for a display-side framebuffer can't be expensive today, a couple of Euros for the extra board space and chip(s). If that stuff takes off, just additional transistors in the all-in-one ASIC.
For the whole screen, that more or less already exists in laptop hardware: "panel self-refresh". HDMI and DiplayPort might need a new extension or something? Is there anything?
The Embedded DisplayPort standard has had the panel self-refresh feature since 2011, and the partial update feature since ~2015. I found a press release from Parade in early 2017 for a TCON supporting the partial refresh feature. I don't think there's anything missing from the ecosystem other than a strong impetus to use those features.
Yes! I work with the Modos team. You’re exactly right - ideally we want region-based refresh hinting. The SDK supports some region based features - we’d like to extend that functionality.
“Traditionally, the [e-paper display] controller used a single-state machine to control the entire panel, with only two states: static and updating,” says Modos cofounder Wenting Zhang. “Caster treats each pixel individually rather than as a whole panel, which allows localized control on the pixels.”
I think there’s a bit of parroting going around but LLMs are predictive and there’s a lot you can inuit a lot about how they behave just on that fact alone. Sure, calling it “token” prediction is oversimplifying things, but stating that, by their nature, LLMs are guessing at the next most likely thing in the scenario (next data structure needing to be coded up, next step in a process, next concept to cover in a paragraph, etc.) is a very useful mental model.
I would challenge the utility of this mental model as again they're not simply tracing a "most likely" path unless your sampling methods are trivially greedy. I don't know of a better way to model it, and I promise I'm not trying to be anal here
Agreed - I picked certain words to be intentionally ambiguous eg “most likely” since it provides an effective intuitive grasp of what’s going on, even if it’s more complicated than that.
Honestly, I think the best way to reason about LLM behavior is to abandon any sort of white-box mental model (where you start from things you “know” about their internal mechanisms). Treat them as a black box, observe their behavior in many situations and over a long period of time, draw conclusions from the patterns you observe and test if your conclusions have predictive weight.
Of course, if someone is predisposed to incuriosity about LLMs and refuses to use them, they won’t be able to participate in that approach. However I don’t think there’s an alternative.
This is precisely what I recommend to people starting out with LLMs: do not start with the architecture, start with their behavior - use them for a while as a black box and then circle back and learn about transformers and cross entropy loss functions and whatever. Bottom-up approaches to learning work well in other areas of computing, but not this - there is nothing in the architecture to suggest the emergent behavior that we see.
This is more or less how I came to the mental model I have that I refer to above. It helps me tremendously in knowing what to expect from every model I’ve used.
They are fundamentally about finding the content that will generate the most revenue. That changes the dynamics quite a bit.
reply