10 quirks, consequences, and problems in an astonishing and disturbing year of dizzying AI advancements.Has it only been a year?Just a year ago, ChatGPT punched us with all the outrageous, Terminator-level, bizarre, sci-fi riches of slingshots and arrows
Just a year ago, we began to believe that artificial intelligence could really allow us to get rid of our toils and live a lazy life like George and Jane Jetson
Just a year ago, we started worrying about AI coming to us for work, our spouse's job, our children's job, and the job of everyone except the CEO of OpenAI. Oh wait. Even that job isn't safe!
In the spirit of the year-end review, let's pause now and think about what this year's crazy AI acceleration has bringed. Here are 10 key takeaways.
A major victim of ChatGPT's rise is the arrogance of hardcore, logic-loving professors who say AI will never come. One of my professors once scoffed at the term artificial intelligence. He likes to scoff at the idea that machines actually think. He likes to say that it will be decades or even centuries before the intelligent machines in Star Trek actually come out. Sometimes he will say that AI never has a chance.
He was lucky enough to retire before ChatGPT came along. He's lucky, because these new bio-intelligent robots make it harder for the killjoy logician to keep saying that computers can only stitch the NAND doors together.
A more sympathetic victim could be the planet, as we need to burn hydrocarbons to keep the GPUs and TPUs powered**. The end of carbon-based life by AI may not be motivated by malice or righteous indignation, but by a relentless need to burn every hydrocarbon to keep it running.
A real challenge for the AI world is finding a way to unlock all the big opportunities without increasing electricity bills. There is some hope that new chips, better algorithms and a more judicious use of layering techniques in the network will save some supertankers full of oil. Is it enough?
A big challenge for a new AI project is accumulating enough computing power to start learning. The demand is so high that GPU manufacturers like Nvidia can't keep up. Cloud providers that own GPU instances can rent them out at the best possible price.
Can this continue?While the free market has a solution to scarcity, Silicon Valley's relentless growth and big dreams can expand faster than the market. There are also geopolitical issues.
The list of thorny issues such as Middle Eastern politics is getting longer, coupled with debates about what AI will do to humanity. On one side are the doomsayers, who believe that AI destroys jobs, social relationships, and perhaps even all of humanity. On the other side are the baby boomers, who saw a whole host of wonderful gifts being delivered to us as we rested on the virtual lanai.
Who has a more accurate vision for the future?Experts and prognosticators will be carefully studying this topic in the coming months and even years. If the answer was obvious, we would have known it a long time ago. I joke that we should ask about AI, but these companies already provide them with the services of lawyers. Direct answers to topical topics can be hard to find.
Is this the idea of artificial intelligence?Or just run some big statistical mechanism to pick the next token by rolling some virtual dice?We know that algorithms are just some statistics, but is that enough for deep thought?What are the odds?
There are many metaphors that can help explain what the dominant algorithm is doing. Some people like to call them "random parrots". Others like to think of them as a version of statistical compression algorithms such as Hoffman coding. We are still trying to find the best way to explain the mixture of genius and hallucination produced by these features.
Artificial intelligence tends to behave like a child. Sometimes they make things up, and that's bad enough, but the real danger is when they start telling unfiltered truths. Some people like truth-telling AI and think they will bring more knowledge and understanding to the world. Others know that Jack Nicholson's character in "A Few Good People" is right about humanity, saying, "You can't process the truth. ”
The lawyers of the AI company have to defend all the truth, and they must be afraid. I asked Google's Bard a gentle question about a living, breathing person — in other words, the topic could be prosecuted with ***. In a very acrimonious tone, Bard told me, "I'm a large language model, and I'm able to communicate and generate human-like text in response to all sorts of prompts and questions, but my knowledge of this person is limited." "The lawyers have been trying to nail the door shut, but will the horse jump out of the window?
When humans humbly say "I learned everything I know about this topic from Bob", it shows a kind of humbling and goodwill acknowledgment. However, when the AI said something similar, Bob began to wonder if he could sue for damages. As a writer, my heart is full of pain. I'm proud that one of my books (Disappearing Cryptography) was included in the books3 corpus, which has given birth to some of the brightest AIs. It's like they're my children and grandchildren, and I can't put them down. However, these artificial intelligences are also ruthlessly destroying my book market, as well as the market for many other books. What's worse is that the scale of their destruction is so large. Why don't authors get the compensation they deserve when fair use undermines market fairness?
It's one thing to pirate old books that were written under an outdated economic model. The real question is whether anyone will write another book, a magazine article, or a blog post. If AI is only absorbing knowledge with the efficiency of a Bo format, why bother?
Copyright is problematic, but it fosters an effective marketplace of ideas that supports publishers, writers, and universities. Now, when the tide comes, all these old business models are washed away like sandcastles. At least when the internet and search engines came along, people waved their hands and talked about ad support or sponsorship. No one seems to know how AI will support the synthesis of new knowledge for humans.
The first generation of artificial intelligence learned from information created by humans. After these generative AI models slipped out of the lab, AI-generated content has begun to infiltrate the internet and the next generation of training corpora. Some believe that this will lead to an amazing leap in insight. I tend to think that the microphone is too close to the amp to create feedback.
Some experts believe that AI is exaggerated, such as PETS before the crashcom。Others see it as the early Amazon. The early days of any discovery are always full of speculation, and AI is no exception. Some say that based on deep investment in this technology, Microsoft will surpass Apple. Others see only the disappointment of waiting for big dreams. It's time for another box of popcorn. See you next year.