I remember a scene in this show which felt like many real meetings I've had in my life. The big hot shot CEO guy pulls everyone into a meeting to share his big idea. The idea? Let's sell a computer that's "twice the speed, half the price!"
...The engineer then rolls his eyes like "yeah no duh". If we could just magically do stuff like that, we would have done it already. Classic management thinking they have an original idea with no understanding of the engineering beneath it all. I thought they would just tell him off and that would be it. I really felt seen in that moment.
The frustrating thing is, they then take pointy haired boss's idea seriously. The rest of the season is spent actually pursuing that dumb, dumb idea... This felt disrespectful, and I stopped watching.
There's a case cited in that paper which does suggest something similar:
> A report in the lay literature describes the case of Claire Sylvia who reported changes in her personality, preferences, and behaviors following a heart and lung transplant at Yale-New Haven hospital in 1988. Following surgery, Sylvia developed a new taste for green peppers and chicken nuggets, foods she previously disliked. As soon as she was released from the hospital, she promptly headed to a Kentucky Fried Chicken to order chicken nuggets. She later met her donor’s family and inquired about his affinity for green peppers. Their response was, “Are you kidding? He loved them… But what he really loved was chicken nuggets” (p. 184, [9]). Sylvia later discovered that at the time of her donor’s death in a motorcycle accident, a container of chicken nuggets was found under his jacket [9].
I haven't read the whole thing, maybe there's something more relevant as well. That report isn't really about accessing the previous persons "memories" but at least claims she adopted a part of their personality. I'd be skeptical about its accuracy without more such reports, however.
A safer assumption would be that our body influences our behavior and tastes, and in turn they are directly affected by changes in our body, like an organ transplant.
A more interesting question regarding the case above would be "what's in our hearth and lungs that affects our perception of capsaicin?".
So if this were true you'd expect people with spine injuries to forget large parts of their lives? Or what is the mechanism to be able to transfer these memories from organs to the brain?
Could you maybe have your harness limit the memory of Claude and then occasionally, when Claude specifically asks for it ("i need to remember something"), you can give Claude the full game history? Most turns, I'll bet it's okay to have a short context and maybe some notes. And then maybe once in a while it's nice to see the full chat history. Wdyt?
Not exactly the same, but kinda: my gen 1 Google Home just got Gemini and it finally delivers on the promise of like 10 years ago! Brought new life to the thing beyond playing music, setting timers, and occasionally asking really basic questions
I think of it more from an information retrieval (i.e. search) perspective.
Imagine the input text as though it were the whole internet and each page is just 1 token. Your job is to build a neural-network Google results page for that mini internet of tokens.
In traditional search, we are given a search query, and we want to find web pages via an intermediate search results page with 10 blue links. Basically, when we're Googling something, we want to know "What web pages are relevant to this given search query?", and then given those links we ask "what do those web pages actually say?" and click on the links to answer our question. In this case, the "Query" is obviously the user search query, the "Key" is one of the ten blue links (usually the title of the page), and the "Value" is the content of the web page that link goes to.
In the attention mechanism, we are given a token and we want to find its meaning when contextualized with other tokens. Basically, we are first trying to answer the question "which other tokens are relevant to this token?", and then given the answer to that we ask "what is the meaning of the original token given these other relevant tokens?" The "Query" is a given token in the input text, the "Key" is another token in the input text, and the "Value" is the final meaning of the original token with that other token in context (in the form of an embedding). For a given token, you can imagine it is as though the attention mechanism "clicked the 10 blue links" of the other most relevant tokens in the input and combined them in some way to figure out the meaning of the original query token (and also you might imagine we ran such a query in parallel for every token in the input text at the same time).
So the self attention mechanism is basically google search but instead of a user query, it's a token in the input, instead of a blue link, it's another token, and instead of a web page, it's meaning.
Read through my comments and those of others in this thread, the way you are thinking here is metaphorical and so disconnected from the actual math as to be unhelpful. It is not that case that you can gain a meaningful understanding of deep networks by metaphor. You actually need to learn some very basic linear algebra.
Heck, attention layers never even see tokens. Even the first self-attention layer sees positional embeddings, but all subsequent attention layers are just seeing complicated embeddings that are a mish-mash of the previous layers' embeddings.
Doge cut muscle sure. They cut the bones too. They sold one of our kidneys on the black market. And then jabbed us in the eyes 3 Stooges style for good measure so we couldn't even see how bad it really was.
We went in for liposuction and buccal fat removal surgery and came out the other side severely disfigured with Maralago face and a hunchback.
That's so generous it's past inaccurate. If they genuinely wanted a liposuction they wouldn't have hired their local pedophile rapist grifter to do it.
What I'm getting from this thread is that people have their own private benchmarks. It's almost a cottage industry. Maybe someone should crowd source those benchmarks, keep them completely secret, and create a new public benchmark of people's private AGI tests. All they should release for a given model is the final average score.
Also still rocking a 13 mini. There are dozens of us! Dozens!
(Also to those who say not enough people wanted a mini phone to be worth producing: I submit the case of Prego chunky pasta sauce. Not many people want a chunky pasta sauce, but you sell a whole lot more pasta sauce in total if you sell both regular and chunky pasta sauce. Malcolm Gladwell has a TED talk about this.)
My pet theory about why the Minis sold poorly is that the 12 Mini was released just a few months after the SE 2; I suspect a lot of the would-be Mini purchasers had just bought an SE 2 instead (not knowing the Mini was just around the corner), and are also not a demographic interested in upgrading their phone every year.
Didn't the SE models sell notoriously bad as well?
There are always a bunch of us who wants a smaller phone, but the sales number indicates that we are the minority.
To some extend I also think it explains the increasingly thin phones. With the increases in screen size, they need to make the phones thinner, otherwise it would feel like a brick in your pocket.
Wasn't eight of the ten best selling phones iPhones? It was even beaten by the iPhone 14 line up, which was only for sale in four out of the twelve months, while the SE was available for all twelve.
You're right that it was good sales figures for a smartphone, but not great for an iPhone.
Apple is like CBS in that regard - shows with viewers that other networks would kill for get canceled at CBS because they expect more. Apple doesn’t get as much ROI on phones that sell in the low ten millions or low single digit percentage of the market. But I do think they could put those phones on a slow refresh cycle and teach e.g. Mini or Plus users to wait for the next release and stack that market together over a few years to be worthwhile to them.
Agree, and as an ex-mini user, I wish this was the world we lived in.
I presume the problem is that iPhones are a lot more expensive to produce than tomato sauce, and it's a lot more difficult to get rid of the ones that people don't buy.
I'm trying to hang on to my 13 mini until next year, when hopefully Apple will have figured out Apple Intelligence and upgrade the phone hardware to run it. But I'm worried my 13 mini won't make it that long. It's starting to struggle with even basic apps and lags just reading news articles. Battery is still at over 80% though
> It's starting to struggle with even basic apps and lags just reading news articles
Not my experience, but I tend to blame the web rather than my phone when things get laggy. I estimate that my 13 is about in the middle of its life, barring unexpected rapid deceleration events.
My battery is at 80% after 2 years. Mine is still on AppleCare+. I always do express replacement so I get a « new » one. Hopefully my battery reach 79% soon.
I kept seeing people mention the 13 mini but I always thought the 12 mini was the last mini. Just looked it up and I see why now, seems like the 13 mini is barely different from the 12 mini.
I had the 13 mini and now the 12 mini and the 13 mini was noticeably better - battery life, camera, screen brightness (big difference when using phone on a bike for navigation)
Speaking of dozens. I’m on a 12 Mini and will be hard pressed to give it up. I also replaced the battery recently and am still quite pleased with its performance!
The added sauce here is they're using it to bias the model during training, not just using steering vectors at inference time (though they do mention that). This is apparently effective at making the intended change in behavior without the lobotomizing side effects that steering vectors can have.
...The engineer then rolls his eyes like "yeah no duh". If we could just magically do stuff like that, we would have done it already. Classic management thinking they have an original idea with no understanding of the engineering beneath it all. I thought they would just tell him off and that would be it. I really felt seen in that moment.
The frustrating thing is, they then take pointy haired boss's idea seriously. The rest of the season is spent actually pursuing that dumb, dumb idea... This felt disrespectful, and I stopped watching.
reply