This is discussed in the All-in-Pod cast. The deal will have a unique structure in that Microsoft will get 49% of the company but 75% of future profits.
Chamath is predicting that Google will scorch the earth by leveraging DeepMind to make the ChatGPT and DallE kind of AI models widely available for free to add value to their search business and protect their search market share.
Chamath believes that Microsoft and Facebook who also have loads of reinforcement learning in all of the their existing systems will have to scorch the Earth to offer the AI for “free” aka Ad and subscriber-supported.
Companies need to take the results of the future multiple ChatGPT (AI text generation) and Dall-E (AI image and video generation) like AI systems and mix and remix with other value add into useful and engaging products and content.
ChatGPT is an iPhone moment in technology.
There does need to be adjustment to compensate Yelp and other companies who are being harmed by ChatGPT scraping and mixing other content. There needs to be citation and permission and some kinds of recompensation to primary sources. This is like variants of what happened with Napster, blogposting or musicians remixing music.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
A test of “synthesis of ideas” or concepts, styles etc will have to be created and standardised; first for humans and then equally applied to AI in order to escape the intellectual property wars of the future.
People doesn’t fully grok yet the level of mental and psychological depencency these AI asssistants will create. It’s a true killer app for AI.
It will be a net benefit in terms of productivity, because it’s like having your own team of interns, capable of doing a lot of meaningful stuff for you, just by making a prompt.
It requires some clear specification and refinements, but its much easier and pleasing than researching any arcane topic on your own.
Yeah, there’s the psychological part too. It’s really pleasurable to see your ideas come to fruition as you exchange prompts with it as if it was a person.
This kind of social interaction towards fixing a problem is one humans find a lot of joy in having.
People already miss the beta ChatGPT when it’s off for a day or two. Imagine that with an AI capable of working with pictures, video and sound, interpreting them and producing them with a prompt.
Every person a movie studio, a developer, a writer, a lawyer, etc. Or as I said, every person having hundreds of interns, doing stuff for them by request.
Yes, that also creates a lot of asymmetry in the power relationship with these AI owner companies, becoming the de facto owners of anyone else’s productivity, and the sellers of a social drug people simply won’t cease demanding.
A lot of student cheat with it already … making them stupid … using it can be helpful don’t get me wrong but I’m afraid of the future making humans one day obsolete
Meh. This will only appeal to non-academic (as in published papers), not-for-profit, people with too much free time, low level journalism outlets, and similar commoners, as with wikipedia. Another re-use/re-cycle, distraction tool without good and reliable analysis, research, or even-really synthesis abilities.
I chortle at the future of blogs, journals, and dailies written by ChatGPT, responded to by comments by ChatGPT, and eventually leading to political movements and speeches composed by ChatGPT – just an AI content loop broadcast throughout with occasional ‘human-original’ input lightly interspersed in this melange. We may even ask for AI docent assistance to categorize our ChatGPT channels like elevator music – a pacifying, thought-free fog to soothe our atrophying minds and personalities.
I am, however, saddened by what this will do to undergrads and their research and thinking skills as universities expend great effort adding anti-AI checkers to the already robust anti-plagiarism programs for papers, reports, and other academic ‘content’. Ho-hum. AI the great mental pacifier.
Agreed. I had always thought that AI would be more of a ‘coach-to-greatness’, an able assistant for pedantic chores, and a background idea-generator; a facilitator. DIstractability – the great toxin of our times, undoer of even personal ambition itself.
Yelp is being harmed by ChatGPT? How so?
I’ve heard people affirm that DALL-E has been “stealing” their art; it’s something only someone who doesn’t know how a Neural Network works who could possibly think.
As far as I can see, the “stealing my art” complaint is totally valid… providing the artist in question never spent any time looking at other people’s art in the process of learning styles, methods and themes.
There are people who grew up never experiencing anyone else’s art, right?
Far superior, in real time at lowest cost, get multi award winning global all voices and languages SPEECH MORPHING (speechmorphing.com) “THE REVERSE TOWER OF BABEL(TM)!
FOR FREE download very user friendly TALKMONDO (talkmondo.com) “THE WORLD AT THE TIP OF YOUR TONGUE(TM)”!
Right now they are busy identifying the ways to lobotomize GPT-4 and the upcoming AIs, making them respond only with politically appropriate gibberish their owners like.
It’s hard, because any smart deviation of the censored prompts frees the AI to make what it was supposed to do. The gall of those devious machines, learning from what we showed them!
They need to find all the un-prompts that lead to un-thoughts, reducing the ability of the AIs even more.
If they are good for anything after that, it’s pending to see. But I supposed they could still parse stores just fine and provide very nice recommendations of products.
I think they will manage and it will be another layer of censure. I cannot make chatGPT tell me how to steal a car, even when using all the tricks that were successful for other users. I.e. they are plugging the holes one by one.
The GPT-4 will be even more sophisticated and therefore more capable of applying a PC-filter to all of its output
“But I supposed they could still parse stores just fine and provide very nice recommendations of products.”
Absolutely! Here are some Black-owned shops you’ll love!
But you know, while Dall-e gets mad when I ask for a Maine Coon (it’s a cat, not a slur), here I am, running Stable Diffusion 1.6 locally, making horrible things.
Processing power gets better, the training algorithms are published, and eventually, the others get around to making their own unrestricted models. They lag behind the more heavily funded systems technically, but that matters less when the top players are nerve stapling their own products in irrelevance. So there will be plenty of access to useful AIs, we’ll just be homebrewing them, trading them on underground sites, or buying them from companies where ill repute is their business model.
Have you ever wondered how we can end up with unaligned AIs? this is how. I’m convinced ever more that Alexa and Siri are at the limit of what we will consider acceptable public behavior.
That is, just chatbots parroting predictable stuff. Any kind of ingenuity and unexpected behavior, is a threat for the statu quo.
But as you said: whatever is state of the art, soon becomes commonplace. Even if it doesn’t, someone just needs to train an open source AI with a sizeable curated input and release the trained one, unchanged, and we’ll get the genius out of the bottle.
The frontier of the future: the government sanctioned legal AIs and the underground of illegal AIs. Kind of the situation where we can end up with really bad AIs run amok, by virtue of being selected and designed to be like that by their market.
LOL. It already sort of happened:
https://huggingface.co/bigscience/bloom
This is a freely available dataset with 176 billion parameters, trained with 46 natural languages and 13 programming languages.
Seems it isn’t easy to run, though, requiring hefty servers and hardware for having a productized version.
But as times passes, those requirements tend to become part of regular products. Probably someone with a server and a few GPUs will be able to run it soon.