The second statement, released in May, was an escalation of both stakes and prestige, a Met Gala of doom. Signed by nearly every CEO of major AI companies and most of the top AI researchers, this statement was only 22 words long. Which really helped the E word: Mitigating the risk of extinction due to AI should be a global priority alongside other societal-scale risks like pandemics and nuclear war.
I’ll serve as The Post’s regular columnist on artificial intelligence for the next year, and the assignment is a relief. For me. I’ve spent months immersing myself in the science, applications, promises, and fears of artificial intelligence, and while I’m increasingly confident that species death is neither imminent nor likely, it’s far less clear what life will be like. I’m grateful that I don’t have to ride the roller coaster alone.
It’s possible that we were at the dawn of an incredible era of tool-making, with the emergence of ChatGPT marking a Kubrickian cut between our mind-bending present and a fantastical future. AI tools are already predicting the spread of infectious diseases, detecting weapons in schools, helping the speechless speak, and reducing energy use. There are brilliant people who say it’s kid stuff. Soon everyone will have a personalized knowledge assistant. The elimination of fatigue and loneliness is coming. Climate change can be mitigated. Rare diseases are commonplace.
Other equally brilliant people go straight to Mad Max: Fury Road. Their scenarios cover everything from an extinction-level virus produced by artificial intelligence to social decay as jobs disappear, inequality becomes permanent, authoritarian states tighten their grip, and meaning is drained from our existence. Before any of this happened, we made great strides in AI-generated porn, fraud, and misinformation.
Both catastrophists and utopians speak a little too insistently. According to the McKinsey Global Institute, this may have something to do with the stakes of $4.4 trillion in estimated annual business benefits from generative AI alone. This is not the Manhattan Project or the space race, when big brains wore government badges. Many of the best AI scientists hold university professorships while cashing checks from the world’s largest tech companies. Now multiply these potential competing interests thanks to the restless mind of Mark Zuckerberg.
When Zuck open-sources the Metas Llama 2 chatbot and invites anyone to play with its guts, is he an optimist democratizing AI (let’s start building!)? Or does he know that the fastest way to reach ChatGPT is for people to hack into his platform for free, even if some of them get Llama 2 to spit out recipes for nuclear weapons? When OpenAI’s Sam Altman puts his name on the 22-word statement on extinction and calls for government intervention, is he worried about bad actors? Or is he hoping for a regulatory oil slick to spray on the companies in his wake? And what better way for everyone to inflate the investment balloon in artificial intelligence than by advertising a technology so powerful that it can enhance progress or destroy humanity? He you know, in the hands of other kids.
The impulse to let this go is strong. Climate change, covid, extremism most people are not looking for more existential uncertainty. Now, keep in mind that AI is moving forward thanks to some of the same companies whose ill-advised social media products have helped drain national reserves of trust and reasoning about things that might be useful right now. It may seem like a Silicon Valley conspiracy to a) dictate the future and b) rob everyone of even a moment’s peace while doing so. No wonder polls show that 82% of voters don’t trust tech executives to self-regulate AI.
But if you look at AI as just the latest tech buzz, a post-cryptocurrency MLM, well, no. Cryptocurrencies are what happens when libertarian financiers get rich with their own assets and turn an interesting but limited technology, blockchain, into a tool for exchanging currencies backed by the full faith and credit of a meme. Cryptocurrency is a bicycle. Artificial intelligence is a high-speed train. He doesn’t need to seduce you with Tom Brady promises.
For more than a decade, personalized song recommendations, improved photos, and easier travel home have been powered by forms of artificial intelligence. What has changed recently is the magnificent abruptness of computing at scale. The number of streets on the planet or songs ever sung seems like a lot of data, but it’s light breakfast for a network powered by a graphics processing unit. GPUs are silicon chips on steroids. Originally designed for gaming, the current generation of GPUs allows machines to process hundreds of billions of different parameters, allowing systems to mimic the multilayered synaptic activation of the human brain.
These neural networks, trained on endless buffets of text and aided by reinforcement learning from human feedback, are the secret sauce. They are what allow you to translate your message into incredibly large numbers and turn it into an elegant linguistic response from the chatbot. Neural networks are still far from reaching biological scale and complexity, but they are already doing something humans can’t do: turning Intel co-founder Gordon Moore’s 1965 prediction into a joke, according to the law by Moore, according to which computing power would double every two years.
Imagine if your brain became 10 times smarter every year for the last ten years, and you were on pace for a 10-fold cumulative increase in intelligence over at least the next five. Add precise recall of everything you’ve learned and the ability to instantly synthesize all that material in any language. You wouldn’t just be the smartest person who ever lived, you would be all the smartest people who ever lived. (Although not the wisest.) This is a plausible trajectory of the largest AI models. This explains how, starting about the middle of the Obama administration, artificial intelligence went from being a precocious child to overcoming many of the supposed barriers between human and machine capabilities. The winners and losers may be ever-changing, but artificial intelligence is likely to creep into most aspects of our lives.
In Silicon Valley there is always the pressure to confuse the inevitable with the instant. The people at OpenAI talk about ChatGPT as their iPhone moment. I bet it looks like that. OpenAI is a young company full of young people. The launch of their product left people speechless. The Netscape moment is a reference that precedes most of them and flatters them less, but it seems more appropriate. Technology is moving fast, but its impact will come in waves. We’ve already noticed slight drops in ChatGPT usage. This certainly doesn’t mean that chatbots are doomed or that artificial intelligence is a fad. Except it’s early. It will take time to understand how people will adapt to these new tools. And viceversa.
That time should not be wasted. The processors will continue processing and we must prepare. That means planning for everything from basic privacy and intellectual property regulations to topsy-turvy labor markets and even Fury Road. Because as far-fetched as the idea of artificial general intelligence—that is, an autonomous system that will one day surpass human capabilities and might consider us, er, superfluous—it’s not impossible. We can also look for the many ways that AI tools can help us solve tough problems that humanity simply hasn’t been able to solve. You were spoiled for choice.
On an individual level, maybe just turn the volume down for a while, okay? That’s the goal here, in this space, to look at AI in a slightly less hysterical way. The story so far has been told by geniuses and scoundrels with conflicting motivations and terrible emotional intelligence. It’s really no surprise that they’re also terrible storytellers. Who starts with extinction?
Let’s start again, this time from creation. All the software we’ve ever used has been designed to work backwards from a result. Its creators wanted to help you find a web page, play a game or use a laptop. Perhaps you’ve noticed that major AI chatbots have come with almost no documentation or user instructions. Not even a piece of clay comes with instructions. That’s what makes this moment unique and worthy of foam finger #1 pride. 1 at the species level. We humans have created a tool for potentially infinite tasks. We are the ones who solve its imperfections and shape its powers.
#Opinion #hate #reasons #reconsider
Image Source : www.washingtonpost.com