Will AI destroy the internet and take your job?

AI generators, chatGPT, stable diffusion, midjourney, open ai and google ai what's the future for you and your job?
Hi I am Ann Reardon, How to Cook That is my youtube channel this week we take a deep dive into the world of AI, open ai, google ai, chat gpt, midjourney, synthesia, runway. What does AI mean for the future of the internet, misinformation and kids learning at school? Is your job at risk?

Feel like binge watching? Here's some more of my videos:

You can send letters & stuff to:
PO Box 202
Chirnside Park 3116
Australia

[lanasinapayen]

As an AI researcher who has to try everyday to explain all of ethical issues with the current popular AI models, thank you for starting with the copyright issue, and for explaining the training process without falling in the trap of presenting the models as actually "intelligent" or having "agency" and making choices like we're in terminator. It's people (companies) making those choices and those choices are not inevitable.

[robertbergeron]

It is staggering to me that the person who is explaining AI in the most practical terms is a cooking instructor. You are amazing.

[Pinkyyyy]

As an artist, AI is incredibly concerning to me since we are actively seeing companies go to AI for artwork in things like animation and book covers rather than hiring or commissioning real artists. art and artists are already so undervalued and underappreciated as it is, this just makes it so much worse. It's especially mind boggling to me since those AIs couldn't even exist without the real artist that they stole the works from.

[lisazimmerman]

I ran into the AI making stuff up problem in my chemistry class this spring. So many students tried using AI to do their honors projects. The more savvy students did rewrite what AI gave them, and submitted solid projects with accurate sources..Other students just submitted the AI product with no effort to check it's information, sources, etc. It was a glaring difference and easy to identify. Definitely going to add a lesson next school year on using AI effectively & pushing our admin to incorporate AI into our academic integrity policy.

[Ea-Nasir_Copper_Co]

A couple of lawyers in the US are on the verge of being disbarred because the junior lawyer used ChatGPT to write a legal document in which the AI _invented_ a bunch of legal cases in support of the plaintiff. Neither the senior lawyer nor the junior lawyer checked the cases to ensure they were real, but the defendant's lawyers certainly did.

[MasterXenex]

"If it doesn't have an answer in its dataset it will just confidently make stuff up" is basically what got me through college and a couple of jobs so I'd argue that's a surprisingly human approach, all things considered 😅

[teomda]

I really appreciate how well thought out this video is: Anne started with visual, more concrete examples, and slowly moved towards more abstract ones, to help viewers understand these concepts. Wonderful!

[ivanvajs]

As someone who does applied AI research I can say that Ann, expectadly, did a great job without being too technical in her description. Keep up the great work! 😊

[jaash]

I'm an artist and writer, and I definitely think Drew Gooden has made the best evaluation of AI art. Art, whether it's drawing, writing, videomaking, photography etc. draws from the human experience and their personal views to create something unique. AI can't have experiences or tell jokes or have any sort of nuance. They don't make original things, and they're just glorified search engines. So even if you could write whole books and movies with chatgpt, I don't care.

The problem with AI are the big corps who are looking at AI like it's their next big moneymaker, and the people who are callous about self-expression and simply think "well my computer can do that now so artists and writers are useless".

Edit to bring some more things up:
I don't think AI is going to replace regular art, except in a corporate sense. Things look bleak, but there's still incredible works releasing today in all forms that are getting the praise they deserve. Don't support blatant cash grabs from corps, like terrible remakes and reboots (ahem Disney). Show that you don't want unoriginality

Second, there is absolutely a difference between AI and commissioning or referencing. The "input" people put into writing prompts for an AI generator is facilitated by other people's work that's been plagiarized, who often do not consent to having their work being fed to AI. If you have a vision for something and don't think you have the skills to see it through, please commission an artist you like the style of. You get your art ethically, and they get paid. Win win for everyone.

Referencing is not simply just copying someone else's work and passing it off as something unique (That's just lying about tracing). It's taking technical cues from them, such as perspective, colors, poses, anatomy, shapes etc. and utilizing them in your own work to enhance your original idea. It's also how we have "Draw this in your style" trends on social media.

[thetableoflegend]

I’m a computer science student and I studied AI last semester. You did a good job breaking down a lot of the issues! One thing I will say is that GPT data sets (although I don’t know about GPT4 specifically) have been scrapped from Wikipedia or Reddit and they continue scrapping all the links it can reach from webpage to webpage as well as some other texts if I’m remembering correctly. This has had many bias issues in the past.

There is also 2 interesting bias issues with datasets themselves.
Most of the data collected in these giant datasets is not labeled, and (in many cases, there are some exceptions) AI need some labeling to begin training. Labeling is time consuming, labor intensive, and sometimes expensive. This has lead to giant datasets where large amounts of it hasn’t actually been checked by a human. These unchecked parts could contain misinformation, disinformation, bias, and it’s all being used to train.
Publically labeled datasets though can also be an issue. An AI needs to have example of words to know what to make from that word. This doesn’t just include nice words. I can’t remember the name of the dataset but there was a public picture database with real humans pictures in it that you could look through. Real people’s images labeled “tweaker” showing a random person who maybe does drugs but certainly isn’t doing any in the picture. They also had categories for slurs. Imagine looking through a dataset and seeing your face under a label for a slur.

Data is a huge and delicate issue. One facial recognition company got most of its data by scraping websites which specifically had no scraping rules, and for awhile their product was being used by the government. Do people realize that every bit of text and every image they post is being taken and used for AI training? Would they change their behaviors if they knew that? What if they were given the option to consent? Is it ethical not to give the option? How much of your information is already out there in a dataset you never consented to?
These are just a fraction of the questions that need to be asked

[janebeckman]

I first encountered AI chat in the shape of an AI "therapist" in 1977. Having been in tech since those days, I have watched the evolution of many tools, AI included. The main problem lies not with AI but in the beliefs of its users. As a tech writer, I have learned to spot patterns (including being fairly reliable in identifying the native language of various writers writing in English). I can generally spot the patterns in AI text, and can trace its evolution, since I've watched it "grow up.". It's still essentially a probability generator.
It has no concept of context. It does not distinguish Joe Smith the physicist from Joe Smith the dead cartoonist, and can mix and match. It literally does not know what "it" refers to. If I switch context, and "it" now refers to a cartoonist's works, it can hallucinate a reference. There is a case where a cybersecurity expert asked what chatGPT knew about him, and it replied with his obituary, and insisted he died in 2017. Even a nonfunctioning URL was furnished as a reference. But since it didn't know that a URL needed to point to collaborating information, it could not show where this information was obtained. But what if you are, say, a government entity who receives information that "you" are dead? You can see where this could go. (BTW, ask it to parse the "you" distinctions in those last two sentences.)
Our biggest dangers are human: the tendency to believe it "thinks" like a human and the tendency to believe its outputs without question.

[stygianoatman]

What worries me most about AI isn't the job replacement, singularity, or copywrite issues. I think the worst part about it is that AI threatens to replace CREATIVITY itself. People can easily fall for the convenience it offers. Why learn to draw when an AI can do it for you? Why learn to write, when an AI can do it for you?

To create art is a journey, it's a wonderful feeling to make things. Getting good at an instrument. Finishing a thoughtful essay or script. Cooking a delicious meal for your family. The quality of that music, writing, cooking, etc. isn't all that AI cannot replicate. It cannot give you a feeling of TRUE satisfaction.

While the outcome of art is fantastic, and I do love a good song, book, meal, etc. those are all just products, and I don't believe we should treat art (or care about it) for what it yields, but the joy it brings to produce. We are human, we were meant to express ourselves, having evolved to communicate using complex language, and making art on cave walls. AI is a quick and easy way to get the product, without the journey and self-expression.

In my opinion, you miss out on a big part of life itself, by letting an AI create something for you. That brain is a muscle, use it or lose it.

Lastly, AI has no love for anything, even if it became advanced to mimic it, there's no hormones responsible for love, a machine cannot feel. If a Grandma bakes you cookies, you know she does it with LOVE. If a robot made you cookies, there is no love behind that. Same goes for making AI create artwork of your pet. An AI feels nothing when it looks at an image of a kitten, puppy, etc... how are people okay with this?

The use of AI can be problematic and insulting to others, but I think it's equally insulting to the self (if not more so). As humans, we should remember we are alive. We DO feel. An AI cannot make something because it loves you, it cannot be your friend, you deprive yourself of true connection by using it. Creation is an expression of love, I believe. A selfless expression (generally). We sing to express ourselves, and it brings people together because we resonate with those feelings. We cook for those we care about. We procreate because we see the beauty of life, and wish for it to thrive and persevere, like it always does.

[anarchy]

It's such a frequent issue people have, that they simply don't understand what they're using. Like the lawyer recently who used ChatGPT and then cited a bunch of imaginary cases.

It's a CHAT bot. It's designed to talk to you and sound convincing. It is not the same as a search engine. You can't go into it expecting the same thing you would from asking a question in a google search. Even on google you shouldn't trust answers without knowing where they come from.

It'd be like having a builder go "I don't get why all these nails didn't go in very well, I hit them with a screwdriver and everything!" It's using a tool for a job it's not designed for, then blaming the tool.

[CoventSynth]

I don't like the "Some jobs will be lost, but others will come" argument. Yes, some will be able to learn these new jobs and work, but for those who lost their job, it may already be too late to learn it, and they will be left behind.
In the end, these people care more about numbers than people, and it shows.

[arladeleon]

I recently watched a news documentary concerning people using some type of system (possibly connected to AI?) to mimic the voice of an acquaintance or relative to scam money from, especially elderly, friends or family. Some lost their life savings. You're right, Ann, not everyone will use these things in honest ways.

[Kyrieru]

One thing to note about Stable Diffusion ai is that it doesn't really learn style, it just learns tags. Ai is incapable of producing poses or compositions which do not exist in its training, and tags are too broad to adequately separate elements from each other. Its why using a tag like "short hair" will suddenly change the entire composition despite pose and composition having nothing to do with hair being short. What you're doing is increasing the influence of the "images of short hair" and all training data close enough to it.

[hollierobinson]

Hands up to Ann for being one of the best content creators out there. The way she tackles all of these real life topics and makes them educational and entertaining is fantastic. I love how much her channel has grown over the years, she’s truly amazing.

[tubthungusbychumbungus]

Its so unnerving going on artstation knowing that by default every post is used to train ai. Because theyre using an opt out system for every individual post, any old posts that haven't been tagged are able to be added to datasets.

[sleepiestmoth]

THANK YOU ANN!! As an artist the excitement around AI bothers me to no end, because it ignores the concerns about copyright in an industry that takes usage rights extremely seriously (to say nothing of the misinformation and mistakes even a beginner artist could avoid). I appreciate you bringing attention to this, more people need to understand why AI has people from artists to computer scientists raising the alarm, and why it isn't paranoid "we're creating the terminator" panic.

[KouryuuProductions]

As a writer, I'm so glad you brought up ChatGBT because it feels like so many people never think of that one, like writing itself is a lesser art than the visuals.