ahalpert
Major Contributor
Specifically they're concerned that people would rather use a chatgpt-type interface instead of Google search for web browsing
is AI gonna take our jobs?
- Thread starter DBP
- Start date
Specifically they're concerned that people would rather use a chatgpt-type interface instead of Google search for web browsing
https://www.yahoo.com/finance/news/apple-stock-nailed-earnings-tim-cook-economy-111807806.html
The stock rebounded. The market cap is $2.42T ... but they will likely cut some spending and investment.
The stock rebounded. The market cap is $2.42T ... but they will likely cut some spending and investment.
NorBro
Major Contributor
That's 80% of their profit, ha...uh ohhh, it's over, we're done, man.
ahalpert
Major Contributor
I'm not sure what you're talking about since Google's searches are already automated. it's not like there's a guy in the backroom reading your search prompts and furiously looking up links to send back to you
ahalpert
Major Contributor
ahalpert
Major Contributor
BTW DLD, what I've read about the implementation of ChatGPT into Bing search is that it synthesizes the information it brings you into written paragraphs, but with footnote annotations showing you where it got the info, and reading it all out next to the search links themselves. so, almost like a wikipedia entry compiled on the fly. basically that sounds like a perfect implementation of the technology into search. I think I'd absolutely use it. and the NYT tech reviewer (Kevin Roose) who tried it says it converted him from Google.
But Google may yet win the race, at least according to an AI researcher whose newsletter I subscribe to. He writes, "Amongst the biggest advantages that Google has over ChatGPT is the multimodal capability of Google’s AI Systems. Simply put, Bard will be far for than a language model. It will integrate language, images, sound and other senses."
https://artificialintelligencemades...-chatgpt?utm_source=substack&utm_medium=email
First mover advantage won't matter that much if Google rolls out a superior product. I'm not backing a horse, just commenting.
In other AI-related news, "This AI Image Fooled Judges and Won a Photography Contest"
https://petapixel.com/2023/02/10/ai-image-fools-judges-and-wins-photography-contest/
And the company Runway now has an AI video generator that transforms videos you supply into a different style (claymation, for example). But honestly, looking at what it can do gives me hope that the problem of generating photorealistic video from scratch will not be so easily solved. Clearly, it's at least an order of magnitude more difficult to accomplish than photographs.
https://medium.com/generative-ai/gen-1-an-amazing-new-generative-ai-video-technology-bb776f5fe8cc
ahalpert
Major Contributor
This is interesting. Apparently Runway's bread and butter commercial offerings are cloud-based rotoscoping and other VFX tools that are supposedly fast, easy, and powerful.
Here's an interview with the team at Colbert at how they use the automated rotoscoping features to meet incredibly tight deadlines on VFX-heavy edits: https://runwayml.com/customers/late-night-show-cbs-uses-Runway/
The pricing is quite reasonable. Some tools are available for free. Moving up to full-quality export and the full assortment of tools will cost you either $12 or $28/month. This is potentially something I could use: https://runwayml.com/pricing/
Here's an interview with the team at Colbert at how they use the automated rotoscoping features to meet incredibly tight deadlines on VFX-heavy edits: https://runwayml.com/customers/late-night-show-cbs-uses-Runway/
The pricing is quite reasonable. Some tools are available for free. Moving up to full-quality export and the full assortment of tools will cost you either $12 or $28/month. This is potentially something I could use: https://runwayml.com/pricing/
A) Text, voice and image integration is obviously a goal but I am not sure everyone will want to go all three on every search.
B) Photo-realistic would have to be cloud based, with the exception of rendering farms.
As to films, I think that if one builds a single image frame via voice or text prompts, one should then be able to subsequently follow that up with a full scene at that "location". In other words, a complete film made from voice/text prompts is entirely possible in the near future.
B) Photo-realistic would have to be cloud based, with the exception of rendering farms.
As to films, I think that if one builds a single image frame via voice or text prompts, one should then be able to subsequently follow that up with a full scene at that "location". In other words, a complete film made from voice/text prompts is entirely possible in the near future.
ahalpert
Major Contributor
Not to return those assets for the search, which you can already basically do. But to process the information in audio and video clips and photographs and report it in the natural language it returns to you when relevant.
I'm sure it would, but I haven't seen a proof of concept to convince me that they're capable of generating photorealistic video in the first place. Maybe it's just around the corner, or maybe not. The runway demo is in some ways not very impressive. Even generated photos are operating at very low resolutions still.
With video game engines like Unreal you can do impressive things but I don't think it actually looks real yet and presumably you need to have a 3D model and a lot of predetermined characteristics about its movement.
ahalpert
Major Contributor
https://twitter.com/fchollet/status/1624817201906868225?t=OmgvbCJeQ5AEgKWyg-Sfjg&s=19
You can read through his brief trials. Long story short - ChatGPT fails to solve unfamiliar puzzles of equal difficulty as puzzles it can solve because it was trained to.
You can read through his brief trials. Long story short - ChatGPT fails to solve unfamiliar puzzles of equal difficulty as puzzles it can solve because it was trained to.
ahalpert
Major Contributor
Here's another AI video editor. Seems like it's fine-tuned for creating internet news-clip type graphics. Intro graphics, sound byte emphasis. And it has a text to image generator component for adding b-roll. The interface is word processor-esque based on an automated transcript. But basically you upload an edited interview to their cloud and then add graphics and b-roll in your web browser using their system.
https://ai.capsule.video/
Unlike Runway, this isn't a product that would be useful for the work I currently do for any regular clients. But obviously videos like these are ubiquitous, and personally i watch many. I'm not crazy about the fake photo b-roll. That's not going to add to my experience as a viewer since videos like these are essentially documentary and I would like to be shown reality.
https://ai.capsule.video/
https://techcrunch.com/2023/02/14/ca...-and-more/amp/
Unlike Runway, this isn't a product that would be useful for the work I currently do for any regular clients. But obviously videos like these are ubiquitous, and personally i watch many. I'm not crazy about the fake photo b-roll. That's not going to add to my experience as a viewer since videos like these are essentially documentary and I would like to be shown reality.
Last edited:
ahalpert
Major Contributor
The reporting about the new version of Bing with incorporated chatbot is fascinating. I can't access the service myself to test it out, but there are too many reports coming from too many places to be ignored.
Bing gets facts and dates wrong and then argues with the user. It says strange, emotional things. (I assume that its creators endowed it with some personality-like characteristics?) When asked about the embarrassing error in Google Bard's launch, it said that Bard misstated the number of countries in the EU because Croatia has recently left the EU. (That wasn't the error, and Croatia is still in the EU.)
More troublingly, it mangled financial data when asked to summarize a financial report of GAP. It misstated things, left things out, made numbers up. Wild.
https://www.digitaltrends.com/comput...-hands-on/?amp
https://www.theverge.com/2023/2/14/23599007/microsoft-bing-ai-mistakes-demo
Bing gets facts and dates wrong and then argues with the user. It says strange, emotional things. (I assume that its creators endowed it with some personality-like characteristics?) When asked about the embarrassing error in Google Bard's launch, it said that Bard misstated the number of countries in the EU because Croatia has recently left the EU. (That wasn't the error, and Croatia is still in the EU.)
More troublingly, it mangled financial data when asked to summarize a financial report of GAP. It misstated things, left things out, made numbers up. Wild.
https://www.digitaltrends.com/comput...-hands-on/?amp
https://www.theverge.com/2023/2/14/23599007/microsoft-bing-ai-mistakes-demo
It's called "teething trouble". The problem with these AI systems is that there's no predecessor to estimate what it will be once fully developed.
(if a frunk flies off a C-8 Corvette, one thinks that GM will solve the problem bu putting higher quality latches ... but here no one knows where exactly this is going and what exactly it will become)
(if a frunk flies off a C-8 Corvette, one thinks that GM will solve the problem bu putting higher quality latches ... but here no one knows where exactly this is going and what exactly it will become)