The A.I. Thread

[Fuzz]

Ya, there is no possible way that's real. Three kids all smiling and eyes open looking toward the camera at the same time? Umpossible.

[Russic]

If we're looking at tell-tale signs, this feels like more bad photoshop than bad AI. I haven't been keeping up with this saga... is the issue that Kate might be missing?

[Firebot]

Originally Posted by jayswin

And it begins. This was predicated years ago that ai generated photos/videos would change the world as we knew it, as society would lose the ability to know what's real or not.

We're officially stepping into that world as we speak.

The above is a poor hastily made photoshop composite, not an AI generation or an example of such. AP has a standard of not using altered / photoshopped photos and this is clearly altered. Pretty much any image you see that is not from a news source has likely been altered or even staged. Photoshop has existed for decades and prior to digital film manipulation was done ever since film was invented.

https://www.boredpanda.com/fake-news...ral-photoshop/

And I mean it when I say it existed since film was invented.

https://hoaxes.org/photo_database/image/baby_adolf/



This should scare people a lot more, bad AI image being believed as real by your average person / FB user.

https://m.facebook.com/story.php/?st...00090644016741

This also is more of a concern.

https://www.bbc.com/news/world-us-canada-68440150

I just hope that people don't suddenly see AI as the birth of deepfakes or deception when it has always existed. It just makes it easier for your average person to create one.

[Cecil Terwilliger]

I just posted about this in another thread. It’s definitely not AI. The palace has come out and said that it was Kate playing around with photo editing. WTF? Do they really think anyone’s gonna believe that?

And if they were editing it because it’s difficult to get a picture of three children smiling at the same time why wouldn’t they release at least one of the originals to prove that it’s legit? This whole thing with Kate is super suspicious.

While I don’t believe all the conspiracy theories that the royal family had her murdered or something it’s really odd how they’re handling this.

[Torture]

Originally Posted by Russic

If we're looking at tell-tale signs, this feels like more bad photoshop than bad AI. I haven't been keeping up with this saga... is the issue that Kate might be missing?

She apparently had some sort of surgery and hasn't been seen in a photo since Christmas aside from a TMZ shot and this one.

[DownInFlames]

I don't really care what Kate's situation is. I do care that manipulated photos and videos, while not a new phenomenon, are going to become increasingly more common and we won't have the time and resources to verify them all.

Misinformation is spreading quickly enough and it's only going to get worse when you can easily create photo "proof" of anything.

[Sr. Mints]

Originally Posted by Russic

If we're looking at tell-tale signs, this feels like more bad photoshop than bad AI. I haven't been keeping up with this saga... is the issue that Kate might be missing?

Photoshopping was required to cover up generations of inbreeding that resulted in lizard skin. I read it on a subreddit.

[Fuzz]

You know how we are all worried about calling customer service, and wanting to get a human? I'm starting to think it won't be long before I prefer to deal with an AI...


These things are coming for all the jobs.

[DownInFlames]

Most CSRs are reading a script off a computer screen anyway, so might as well make them bots.

We'll need something like UBI sooner or later unless someone can find new ways to keep us more productive than machines.

The rich and powerful will say they don't want to give away free money, but they should look at it as investing in not getting a pitchfork to the face.

[Bill Bumface]

Originally Posted by DownInFlames

Most CSRs are reading a script off a computer screen anyway, so might as well make them bots.

We'll need something like UBI sooner or later unless someone can find new ways to keep us more productive than machines.

The rich and powerful will say they don't want to give away free money, but they should look at it as investing in not getting a pitchfork to the face.

I wish we'd get more creative with UBI.

Having a purpose is important. Sitting on the couch playing video games and cashing a monthly UBI cheque probably does make for the happiest person. What if instead we took that money and put it to all of the work that isn't getting done that would be a benefit to all, and provide people with some pride and purpose?

Spitballing some ideas:

- Paying a bunch of people to plant flowers, pick up litter, weed and maintain public park spaces
- Pay youth/rec sports coaches and art teachers so that we can drop the fees on those programs and expand the reach
- More public art everywhere
- Assigned community police officers walking around neighbourhoods checking in with citizens and building relationships and gathering data

I'm sure there are a billion things I'm thinking of that have relatively few raw material inputs that still produce good for society that could be dreamed up.

[activeStick]

Originally Posted by Fuzz

You know how we are all worried about calling customer service, and wanting to get a human? I'm starting to think it won't be long before I prefer to deal with an AI...


These things are coming for all the jobs.

I love the advancements in AI! So darn exciting. The firm I work for recently stood up a Generative AI studio in our Toronto office intended to serve as an innovation hub for AI specifically. Since we made it a corporate goal last year to get everyone versed in AI, our clients have all been asking all our client account leads the expected questions around what we're doing to deliver in this new environment and what efficiencies they can expect from us. Very exciting.

[Russic]

Originally Posted by Fuzz

You know how we are all worried about calling customer service, and wanting to get a human? I'm starting to think it won't be long before I prefer to deal with an AI...


These things are coming for all the jobs.

Pretty sure I dealt with my first "smart" chatbot the other day. The only real tell was that I hate the company in question and the customer service whatever on the other end of the chat was beyond helpful and crystal clear with all information. At the end I was met with a survey that had several questions along the lines of "I'd definitely like to talk with this agent again" and "this agent was much more helpful than previous times I've contacted <company>".

As for the robot in the video above, I feel like that's one of the big logical leaps that few are thinking about. A year ago training a robot to clean up your house would feel overly complicated... now I'm sure it's still very complicated, but I can see how it might work. "All books go here... kids toys in this bin..." kind of stuff should technically be "programmable" just by talking to it.

[wireframe]

Originally Posted by Bill Bumface

I wish we'd get more creative with UBI.

Having a purpose is important. Sitting on the couch playing video games and cashing a monthly UBI cheque probably does make for the happiest person. What if instead we took that money and put it to all of the work that isn't getting done that would be a benefit to all, and provide people with some pride and purpose?

Spitballing some ideas:

- Paying a bunch of people to plant flowers, pick up litter, weed and maintain public park spaces
- Pay youth/rec sports coaches and art teachers so that we can drop the fees on those programs and expand the reach
- More public art everywhere
- Assigned community police officers walking around neighbourhoods checking in with citizens and building relationships and gathering data

I'm sure there are a billion things I'm thinking of that have relatively few raw material inputs that still produce good for society that could be dreamed up.

You're touching on an established idea called universal job guarantee. I'm a huge supporter of this idea for the exact reasons you mentioned - there's lots of stuff that needs doing and no current financial incentive to do it. It also sets a minimum wage; current minimum wage is $0 if you're unemployed.


My radical idea is that we can do this with a new branch of the military that is organized close to the community level.

[chemgear]

[Firebot]

I just got access to Claude 3 which is finally available in Canada it appears as of today. I've been using Claude 3 via POE and perplexity to bypass the Canadian restrictions. It's amazing to see it available in Canada.

It has 3 models, Haiku (meant to be a chatbot or simpler tasks), Sonnet (middle model) and Opus (top model, highest resources and slower)

Sonnet is currently free to use and performs nearly as well as GPT 4 from my tests and in some ways even better. I highly recommend it. Opus is actually better than GPT4 on benchmarks, and does coding even more efficiently without as many errors. Claude 3 is also a lot more relaxed than its predecessor which was notoriously and hilariously censored. The API offers 200K token context as well. Sonnet /Opus also seems more fresh and provides more detailed answers (GPT4 has a lot of nuances and issues with laziness that you can easily spot when using it extensively daily). I will be testing it extensively now that I have direct access.

Grok also released it's new model yesterday, available to Twitter premium subscribers, it appears to perform quite well, nearly on par with GPT 4. And it appears to be uncensored as well (see videos on Matthew Berman's tests below)

I thought this would also be a good time to give some additional AI related information and resources for anyone looking for more tutorials or want to get deeper into the subject, want to be in the know for newest updates.

The best channel for AI news

https://www.youtube.com/@mreflow

2nd one (don't let the obvious bait titles fool you, it's a great channel).

https://www.youtube.com/@WesRoth

If looking for tutorials on AI image generation these are your best options

https://www.youtube.com/@sebastiankamph

https://www.youtube.com/@OlivioSarikas

Local LLM fine tuning, tutorials (if you want to know how to create your own uncensored AI GF chatbot or roleplay bot, this is your best source to know this)

https://www.youtube.com/@Aitrepreneur

These 2 are more advanced by programmers who dive deeper and do extensive testing of all LLMs, very valuable resources if looking to build AI aided programs, agents, etc.

https://www.youtube.com/@matthew_berman

https://www.youtube.com/@AllAboutAI

For music AI, by far the best is Suno, I have access to V3 and the music generated is absolutely mind blowing https://www.suno.ai/

For voice generation, use ElevenLabs https://elevenlabs.io/

For image generation, your big 3 are:
Midjourney (closed source, censored, subscription, amazing photorealism and best quality images overall)
Dalle 3 (neutered photorealism, most coherent model, can be used for free via Bing / Copilot or as part of a chatgpt subscription)
Stable Diffusion (models were released to public and can be run locally, and likely the source of all NSFW AI imaged on the internet right now, SD3 is about to be released soon which is close to Dalle 3 in text generation)

For video generation, until Sora is released, the best text to video and image to video tools are

Runway https://runwayml.com/
Pika Labs https://pika.art/

Both are good, but I wouldn't call commercially viable at the moment.

Hopefully some of this helps for anyone who's curious.

[Russic]

Thanks so much for the breakdown. What are the usage limits on Sonnet and Opus? It still says unavailable in Canada to me. Perhaps it's rolling out slowly?

Originally Posted by Firebot

...

For music AI, by far the best is Suno, I have access to V3 and the music generated is absolutely mind blowing https://www.suno.ai/

Whoa. Holy hell you weren't kidding. Whether this is your genre or not, I wouldn't be able to pick this out from a stack of pop songs and identify it as AI.

[edslunch]

Originally Posted by DownInFlames

The rich and powerful will say they don't want to give away free money, but they should look at it as investing in not getting a pitchfork to the face.

They should be worried about people not being able to afford their stuff

[Firebot]

Originally Posted by Russic

Thanks so much for the breakdown. What are the usage limits on Sonnet and Opus? It still says unavailable in Canada to me. Perhaps it's rolling out slowly?



Whoa. Holy hell you weren't kidding. Whether this is your genre or not, I wouldn't be able to pick this out from a stack of pop songs and identify it as AI.

I had the email and was able to sign up, but couldn't with another account, I also couldn't subscribe to Claude Pro, perhaps it was prematurely sent and reverted back. You can sign up using a VPN, or use Claude 3 via POE or perplexity. I may try out perplexity as it has unlimited usage of Opus right now.

The free version of Sonnet on Claude is about 20 messages every 3 hours or something (you hit the limit quick). Sonnet is surprisingly good on coding as well. Apparently the limit for Claude on paid subscription in chat is 100~ messages if they are short and they refresh every 8 hours.

[Huntingwhale]

That Suno.AI site is...amazing. Mind blowing stuff. My song description was make a song about my grandmother's fat ass and how she cooked bacon all the time and it made this.

https://app.suno.ai/song/8d14fc36-2c...d-908ba6b80bd0

Like...WTF. Amazing

[Bagor]

An AI tool tested by an NHS hospital trust successfully identified tiny signs of breast cancer in 11 women which had been missed by human doctors.

The tool, called Mia, was piloted alongside NHS clinicians and analysed the mammograms of over 10,000 women.

Most of them were cancer-free, but it successfully flagged all of those with symptoms, as well as an extra 11 the doctors did not identify.

https://www.bbc.com/news/technology-68607059