Plato wrote about a lot of these kinds of words in the dialogue Cratylus. The dialogue is about more than just etymology, of course, but it’s worth looking at if the subject is interesting to you.
Very well put. I've also heard it said like this, "abusing the truth is the easiest way to lie".
The pattern is 1. take something that's true then 2. use it to imply the Truth. Step 3. is if anyone doubts 2, challenge them to disprove 1. Even if you don't deny that 1 is true and just challenging that 2 doesn't follow from 1, the waters have already been muddied in the eyes of many onlookers.
It's silly that this works but there's so much over-communication now that simplicity is very effective. People often repeat the fallacies verbatim too.
Has anyone seen the corridor crew's green screen ML project? They're on YT and they trained a model by using 3d objects, which have perfect transparency, and then adding post facto green/blue screens. Surprisingly, very little training data was needed as the data that was used was perfect by construction. I think right now it's the best plugin of its kind in the world, and they built the prototype in like a weekend.
What I think this illustrates very clearly is this type of technology responds very well to good data, and that to have good data you need to have a clear goal.
This is why it seems that alignment for a generalized, chat-style AI is a very hard problem, perhaps impossible. You can't align it to solve a certain kind of problem and keep it general to any question. The two goals are in conflict with each other.
I think it was Sam Altman himself who said (I don't remember when or where, sorry) that the reason he was so confident in this technology was he noticed the gigantic leaps it made in certain areas in response to even a small amount of training.
(This is why LLMs are so strong at coding, because it's overrepresented in training data. My guess is that if you ask a frontier model about makeup, you will see it repeat cosmetic company's copy rather than getting a chemistry lesson.)
This makes perfect sense but it does seem to kind of be at odds with the concept of a general AI whose job is simply to be smart at any goal. How do you train for any goal?
I guess in a way the AI makers suffer from the same problem that we humans do. We would all love a solution to everything, but to do that you need to define the goal. I'm not sure if that's a tractable problem.
I'm guessing the future is more geared towards specialized AI that are very good at solving the problem they were trained to do, and a human who knows how to breakdown a larger goal into smaller ones by composing the solution out of these models. This also seems like the more efficient solution as well, and better aligned with other goals like privacy and safeguarding of IP.
> This is why it seems that alignment for a generalized, chat-style AI is a very hard problem, perhaps impossible. You can't align it to solve a certain kind of problem and keep it general to any question. The two goals are in conflict with each other.
I think this problem is going to be solved soon (hopefully), check out for the Persimmon model[0] from Humans&. They train it to mimic humans, it may seems bad but it could be _really_ useful to train an AI to be aligned to humans and understand their goals really well since they can use Persimmon to create a "fake human" following a defined goal that their big AI model will learn to estimate.
It's still early but I think this is what they are heading toward.
I think we should start training models to guess intentions. That's the whole game. e.g. If I dislike a particular short form video, what are the specific intentions I am conveying to the recommender system, (general category I don't like, the time at which it was shown, the combination of feed in which it came), these are all intentions I am trying to convey. In websites google analytics helps with clickstream tracking (or clickstream analysis, user journey mapping or behavioral tracking.) — the sequence of clicks and pages you move through is your clickstream, and analyzing it to infer intent. The same thing will need to happen in chatbots. Can a chatbot infer the specific instance of the problem I want to solve?
> This is why LLMs are so strong at coding, because it's overrepresented in training data. My guess is that if you ask a frontier model about makeup, you will see it repeat cosmetic company's copy rather than getting a chemistry lesson.)
ChatGPTs response to the question “I want to learn about makeup”, gave me an overview of what makeup does, how it affects perceived structure, complexion, evenness, geometry, texture, etc.
When you then ask “the chemistry of makeup”, it goes into interesting breadth and depth without seeming like proprietary information. I do t get “corporate PR or marketing” vibes.
The only problem is, there’s a lot of money tied up and openAi and Anthropic, who are incentivised to convince the world that the general approach is the money making one.
Not that he got it all right or anything like that but it’s amazing how much of the current macro discussion around AI capabilities is covered in the Foundation series which was published over many years, 1942 to 1993.
If I were a business, I wouldn't want my proprietary business information into a hands of a competitor. But CEOs are racing to do it, and (at best) relying in flimsy contractual guarantees they have no means to verify.
What really matters is the claim that Apple never raised those issues in the lawsuit with them. OpenAI claims the emails prove this, but to me all the emails prove is Apple sent an email that said “please see the attached letters”. We don’t know what is in the attachments but I imagine those were the relevant issues they raised.
Moreover OpenAI says, “[Apple] now admits that Apple employees reached out to him and asked for his help to locate this information”, but the transcript they cite seems to me to show the opposite? It’s Chang Liu asking for someone to help him copy data. The part where Apple employees reach out is later and it’s not for copying data to a hard drive… it’s basically “who should I talk to about this project?” Which makes sense because when he was contacted it was after he left (so he wouldn’t have access to data)
Can someone please tell me what am I missing? None of this makes sense with what they’re saying. The fact that the lawyer emailed the wrong person is moot if they caught it and resolved it days later.
The timeline OpenAI wants to present looks like this to me:
- Leaving employee (LE) needs a USB stick to copy data off their Mac to give to their Manager (MA) so someone at Apple still has access
- After LE left, MA has access to LE Mac and iCloud to still access some data
- At some point, LE asks MA to sign out off LE's iMessage account on that Mac as they suspect they'll get OpenAI-related messages soon and don't want Apple/MA to see them
- Eventually, MA deletes the Apple files in LE's iCloud
Granted, Apple does do this weird thing where you're invited to use your personal iCloud (/Apple) account for work-related things (I suspect because they're just bad at supporting multiple accounts simultaneously). As part of the deal, your iCloud account gets upgraded during your employment.
But they're also quite explicit and repeat the directive (including during the exit interview) about removing Apple files from your account when you leave, and logging out of iCloud from the Apple devices you leave behind.
Also:
> At some point, LE asks MA to sign out off LE's iMessage account
You can sign out remotely, so this would've been stupid of everyone involved.
Then again, this whole situation demonstrates a level of stupidity from supposedly competent people that I wish I could say was unbelievable. Add "Making a public post about a topic under litigation" under that category too.
From what I saw, it's been leapfrogging for the past few weeks. Even if Apple is currently behind them, I don't think it's particularly fair to say it's "fallen to" them.
Funny to hear that even Apple themself struggle with that. In the past it's been such a PITA for us to decide whether employees can use private account or shall use company apple ids. As we have a pretty lax policy, most of them use their private accounts to have iCloud sync or to simply access purchased apps but having a simple way for dual account usage would be too easy i guess.
I am actually confused why creating an email john.doe@apple.com and separated employees work only iCloud on the email is such a problem and they want people to use personal iCloud? Not even corporate stuck in 90s does that.
I wonder how deep this integration goes. I’d never want to use my personal accounts for proprietary work items. Just seems like a recipe for disaster. Also, removing a latop from iCloud doesn’t delete the imessage data already on the device.
I was a bit confused too initially, but it looks like Chang was asking for the data to be copied _off his iCloud_ by the existing employee on his final day of employment. And then everything after that is just the existing employee asking for help on things Chang had prior knowledge of
Why in the world would a high-level Apple employee not sign out of his personal iCloud on his work computer, and then also use his personal iCloud for data from his new company?
reply