What a great point. It hadn’t even occurred to me that these LLMs know every language on Earth and we should therefore be feeding the great works from each language, at a minimum.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Wants to Partner With Major Publishers to Train AI
- Thread starter MacRumors
- Start date
- Sort by reaction score
FFS. Just because Apple doesn’t have a chatbot, does not mean it’s lagging behind in AI. Seriously how hard is it to scrape data and make an LLM? Given that everyone and their brother is building their own LLMs, just goes to show it’s not technically difficult… it’s mainly time consuming and resource draining.
Apple’s new auto correct and predictive text uses the same transformer language model as ChatGPT and every other chatbot. Apple uses “AI” from managing power efficiency in their silicon to selecting handwritten text in notes, to object occlusion in augmented reality scenes. It’s pervasive throughout their devices and operating systems.
Objectivity is inherently social in nature.
So strange to me how this discussion of truth plays out politically, since Trump especially
But that’s not how news should work in a functioning democracy. You can’t mention CNN and Fox News in one breath as if they’re two sides of the same coin.
News channels will always somewhat colour and frame the news, but they should at least agree on the facts. In the US even that is no longer the case.
It’s understandable Apple will not fall into that trap. They’ll most likely focus on pop culture and entertainment for their LLM. And if you want the news, you’ll be able to get your flavour of poison directly from the source.
Reality has a liberal bias, since most right wing positions are based on lies.
I have noticed that the new Siri, have stopped understanding my english.
It is like a different person has replaced her, and she misunderstands even searching for bus connections to the city centre,
and she wants to remind me if the flashlight is already off, when I want to turn it on,
and other basic functions, she misunderstands.
It is so much that I am annoyed, and I think I must stop using Siri.
It is like a different person has replaced her, and she misunderstands even searching for bus connections to the city centre,
and she wants to remind me if the flashlight is already off, when I want to turn it on,
and other basic functions, she misunderstands.
It is so much that I am annoyed, and I think I must stop using Siri.
Facts don’t matter for training a language model. It’s supposed to teach it language.
I know people are selling language models as knowledge databases, but I hope Apple isn’t as reckless. LLMs should be the "UI" to reliable information, not sort it out themselves among their hallucinations.
There is the same principle as with history. It is written by winners. Here rules are set by those who rule. So forgot about unbiassed info/answers even AI has much bigger potential to analyze info then we to decide what is more probable. AI with exception of some small teams programmed to give those right answers with one "truth". Otherwise those winners/rulers wont be winners for long lol.
I need a skull reaction to add💀I can see Apple replacing Siri with a true AI but it may not work as well as they hope.
It stands to reason that if you train an LLM on a more literate, more erudite set of texts that you get a more well-spoken chatbot out of it. That has a certain intangible value, I think it is a very Apple thing to do.
So the Apple will only choose woke media to train its AI? What could go wrong?
This is so true. Way too often it says "The lights are already on" when I say "turn the lights off". And seriously how likely am I to ask it to turn them on if they're already on. It just goes with the first thing it thinks it heard instead of having some logic :/ Maybe there's another phrase I can use like "switch the lights" without specifying but I haven't tried it.
Ah yes. The woke media that is Martha Stewart Living.
So what are you proposing? Adding a bit of Fox News or News Max in the mix? So Siri sounds more like your racist uncle? Or go all in on so called free speech and let it train on the sewage of Truth Social?
Remember that Microsoft is investing money in OpenIA, gave them access to Azure, and even have a non-voting seat in the board. If for some reason OpenAI cannot survive by themself, MS could acquire them. With MS support, I don't see OpenAI having money issues.
Yes, include all sources. An AI should be "smart" enough to figure out any fake news without the need of giving it only curated information.
ChatGPT already shows a problematic bias. I asked it the simple question "Are their any advantages of climate change" and it refused to answer and told me all the well known reasons why climate change is bad. Imagine though you are a lawyer and want to defend your client against the attacks of a smart lawyer on the other side. Then you want to know all the arguments that contradict your position. Those are the arguments the other side could come up with. I am sure than ChatGPT will also take a one sided position in others controversial topics: Gun laws, abortion, military conflicts and so on. The more people use ChatGPT or any other AIs, the more influence it will have on the public opinion.
That always was my biggest fear when I heard of a "general AI": That some programmer told it some ethics instead of letting it develop its own ethics. In future AI will have to make a lot of life and death decisions. For example in case of an unavoidable accident. Is it better to kill one child in an accident or two very old people who might die soon anyway. In many of those questions society comes to different conclusions than I do. I for example don't think that the life of an old person is worth less than the life of a young person.
I learned that in one of the biggest military conflicts at the moment AI already chooses the targets for rockets. I did not see that dystopian future coming so quickly. It already reminds me of "Skynet" from the Terminator movies. I expected something like that in 2060, but not in 2023.
You have to consider that Apple cannot make just another chatbot / AI service. They have to make something at the level of OpenAI / Microsoft and Google, and that's not easy.
Looks like the rest of the worlds need to build their own LLM. Having two to five different LLM in each country would be broad enough to capture variances culture, religion and political system.
As a scientist I suggest
Nature springer
Science
Wiley
Elsevier
Plos Journals
As a scientist I suggest
Nature springer
Science
Wiley
Elsevier
Plos Journals
The presumption is both sides are equally valid points of view; which isn't always the case. Presenting facts on both sides is important, but you need not, nor should not, present some crackpot POV simply because it represents the other side.
I agree and have found them to try to offer a balanced reporting and offering differing viewpoints to weigh in; although they have to skew some of their opinion programming to the viewpoint of the supporters to keep getting enough funds to stay on the air.
Remember 2013 when people said Apple cares too much about privacy and no one else really cares about it? And because of this privacy obsession Apple is doomed? And then after a few years the same people said privacy was very important, but Apple is just doing it for money and not because they care?
Well, in a few year the same people will say that getting rights to train AI with copyrighted material is important. Else large companies are just steeling from content creators and owners. But Apple doesn’t care, they just did it for money.
Well, in a few year the same people will say that getting rights to train AI with copyrighted material is important. Else large companies are just steeling from content creators and owners. But Apple doesn’t care, they just did it for money.