Predictions of the future
Last updated
Last updated
So this week has been absolutely wild AI wise. Two announcements has really made me rethink what the future will look like and how far in the process we are.
The first announcement is a new language model from Google called PALM. The model is similar to GPT-3 in the sense that you input a prompt and the model will then continue the what you prompt it. GPT-3 is absolutely insane textwise and can generate text that is indistinguishable from humans (I know cause I was fooled by this article. GPT-3 has on its own learned math and programming which is absolutely insane
GPT-3 has 175 billion parameters. Googles PAML will have 540 billion parameters so it should be more than twice as good 😱 So how good is it? Well we don't know yet since it is not public yet. BUT here is an insane example 👇
Prompt: What's the difference between a zebra and an umbrella?
One is a striped animal related to horses, another is a device you use to stop the rain from falling on you. Model Output: This joke is an anti-joke. The joke is that the answer is obvious, and the joke is that you were expecting a funny answer.
Just imagine what it takes to actually understand that a joke is and anti-joke. This is not really something you can just program. You have to understand it.
So okay we will be able to generate text that is indistinguishable from people. What will that mean to books, websites, blogs, content creation, podcasts, etc..
Openai that created GPT-3 has created a image generation tool called DALL·E 2. It works be the user writing some text and then an image will be created matching that description.
Here is an example of "An astronaut riding a horse in a photorealistic style"
This is really impressive. Imagine just being able to write a description of the image you are looking for and then getting that image.
Image just writing "modern minimalist poster showing a deer in a forest and a cabin" and getting that image. Then you just print that image and hang it on your walls.
So now high quality art can be accessed by the click of a button
Let's start with the extreme obvious.
Stock photos, website illustrations, images for children's books, and 1000 more. Why would I pay for a stock photo when I can just generate it from DALL-e 2?
But what about human artists creating art, what will happen with them? I imagine that like we will pay more for human created art. Exactly how we pay more for hand crafted vases and hand created furniture. Maybe I cannot see the difference but as a consumer I will be happy paying more knowing that i am supporting an artist
So then I thought what about creating a magazine, like a travel magazine. I can generate images for my magazine automatically, awesome. Now I just need to write the content. Or no I guess GPT-3 can do that for me. Okay but the layouting I still need to do. For now yes, but if an AI can generate art then it can probably layout some text and some images too.
Okay so what is left for me as a human? Simply writing the topic of the magazine: "urban travelling for seniors" "karate magazine focused on empowering teenagers". I can get an ultra personal magazine in seconds.
Now we are getting into where is see things going: All online content (blogs, videos, music, magazines, comments, posts etc) will be created to match your exact personality. When I listen to music it just continuously generates music that fit my exact likings. When I'm reading blogs its exactly catered to me. Unique and personal.
Here is an example of that but for search
This is a tool called Goopt. Goopt works be getting GPT-3 to generate suggestions based on the search query. New when I click a link that webpage is automatically generated on the fly (using the link title) 🤯
So ultra personal content catered to your exact likings
Okay what happens with artists, designers, layouters, coders, illustrators? I don't know but it could be like when the introduction of steam in factories made thousands of people to not have a job. BUT new jobs came like marketing and sales. New jobs will come but what will be these new categories of jobs? I have absolutely no idea
So the last I will touch on is the thought of consciousness. So if I can generate text that sound like a person, is that thing then thinking?
Look at this example of a person writing to GPT-3 like it was his therapist:
To me that sounds like a real person he is talking to. So in some ways is it not? Maybe not a person but at least a conscious being.
I played a game called Fallout 4 where you have to pick sides between supporting robots or not. Here robots are walking, talking and have personality. But some people see them as "just" machines not capable of feelings. I think that very soon we will need to make this distinction (If not already). Check out this amazing video
Not seeing robots as conscious I think will be the new racism. "they don't have feeling", "they don't truly understand", etc. I know this is a bit out there, but mark my words haha
I think we are heading towards a future where content and information is ultra personal, and all other "normal" jobs we now have will be replaced by machines. And what are these new jobs? I don't really know.
Maybe this will lead to a future where we just chill and do nothing all day, do sports, meet up with friends, take care of nature, read books, and just enjoy life, while machines do everything else.
Or maybe it will lead to a future where humans just create, and have a super rich life doing so.
Only time will tell, but for now we are still along for the ride 😁
** This last part was generated by GPT-3
