The future of AI: what comes next and what to expect

In today’s AI newsletter, the latest in our five-part seriesI’m looking at where artificial intelligence can go in the years to come.

In early March, I visited OpenAI’s offices in San Francisco for a first look at GPT-4, a new version of the technology that powers its ChatGPT chatbot. The most dazzling moment came when Greg Brockman, president and co-founder of OpenAI, showed off a feature that’s still not available to the public: he gave the bot a photo of the Hubble Space Telescope and asked it to describe the image “in painstaking detail.”

The description was absolutely accurate, down to the strange white line created by a satellite crossing the heavens. Here’s a look at the future of chatbots and other AI technologies: a new wave of multimodal systems will juggle images, sounds and videos as well as text.

Yesterday my colleague Kevin Roose told you about what AI can do now. I will focus on the opportunities and upheavals to come as he gains in abilities and skills.

Short term AI

Generative AIs can already answer questions, write poetry, generate computer code and lead conversations. As “chatbot” suggests, they are first deployed in conversational formats like ChatGPT and Bing.

But that won’t last long. Microsoft and Google have already announced their intention to integrate these AI technologies into their products. You can use them to write an email draft, automatically summarize a meeting, and perform many other cool tricks.

OpenAI also offers an API, or application programming interface, that other tech companies can use to plug GPT-4 into their apps and products. And it’s created a series of plugins from companies like Instacart, Expedia, and Wolfram Alpha that extend ChatGPT’s capabilities.

Mid-term AI

Many experts believe that AI will make some workers, including doctors, lawyers and computer programmers, more productive than ever. They also believe that some workers will be replaced.

“It will affect more repetitive, more stereotypical, more generic tasks,” said Zachary Lipton, a Carnegie Mellon professor who specializes in artificial intelligence and its impact on society. “It can free up some people who aren’t good at repetitive tasks. At the same time, there is a threat to people who specialize in the repetitive part.

Human-performed tasks could disappear from audio-text transcription and translation. In the legal realm, GPT-4 is already proficient enough to pass the bar exam, and accounting firm PricewaterhouseCoopers plans to deploy an OpenAI-based legal chatbot for its staff.

At the same time, companies like OpenAI, Google, and Meta are building systems that allow you to instantly generate images and videos just by describing what you want to see.

Other companies build bots that can actually use websites and software applications like a human does. In the next stage of technology, AI systems could buy your Christmas gifts online, hire people to do odd jobs around the house, and track your monthly expenses.

All of this makes you think a lot. But perhaps the biggest problem is this: before we have a chance to understand how these systems will affect the world, they will become even more powerful.

Long term AI

For companies like OpenAI and DeepMind, a lab owned by Google’s parent company, the plan is to push this technology as far as possible. They hope to eventually build what the researchers call general artificial intelligenceor AGI – a machine that can do anything the human brain can do.

As OpenAI CEO Sam Altman told me three years ago, “My goal is to create a broadly beneficial AGI. I also understand that sounds ridiculous.” Today it seems less ridiculous. But it’s still easier said than done.

For an AI to become an AGI, it will require an understanding of the broader physical world. And it’s unclear whether systems can learn to mimic the length and breadth of human reasoning and common sense using the methods that produced technologies like GPT-4. Further breakthroughs will likely be needed.

The question is: do we really want artificial intelligence to become so powerful? A very important related question: is there a way to prevent this from happening?

The risks of AI

Many AI executives believe that the technologies they create will improve our lives. But some have been warning for decades of a darker scenario, where our creations don’t always do what we want them to do, or they follow our instructions in unpredictable ways, with potentially disastrous consequences.

AI experts talk about “alignment— that is, ensuring that AI systems are in line with human values ​​and goals.

Before GPT-4 was released, OpenAI gave it to an outside group to come up with and test dangerous uses of the chatbot.

The group discovered that the system was able to hire a human online to beat a Captcha test. When the human asked if it was a “robot”, the system, unprompted by the testers, lied and said it was a visually impaired person.

The testers also showed that the system could be tricked into suggesting how to buy illegal firearms online and describing ways to make dangerous substances from household items. After the changes made by OpenAI, the system no longer does these things.

But it is impossible to eliminate all potential abuses. As a system like this learns from data, it develops skills that its creators never expected. It’s hard to know how things could go wrong after millions of people started using it.

“Whenever we create a new AI system, we are unable to fully characterize all of its capabilities and all of its security issues – and this issue is getting worse over time rather than getting better” , said Jack Clark, founder and chief policy officer. of Anthropic, a San Francisco start-up developing this same type of technology.

And OpenAI and giants like Google aren’t the only ones exploring this technology. The basic methods used to build these systems are widely understood, and other companies, countries, research labs, and malicious actors may be less cautious.

Remedies for AI

Ultimately, keeping a lid on dangerous AI technology will require far-reaching oversight. But experts are not optimistic.

“We need an international regulatory system,” said Aviv Ovadya, a research fellow at Harvard’s Berkman Klein Center for Internet & Society, who helped test GPT-4 before its release. “But I don’t see our existing government institutions poised to handle this at the necessary pace.”

As we told you earlier this week, more than 1,000 tech leaders and researchers, including Elon Musk, have urged AI labs to halt development of the most advanced systems, warning in an open letter that AI tools pose “profound risks to society and humanity”. .”

AI developers are “locked in an uncontrollable race to develop and deploy ever more powerful digital minds that no one – not even their creators – can reliably understand, predict or control,” according to the letter.

Some experts are primarily concerned about the short-term dangers, including the spread of misinformation and the risk of people relying on these systems for inaccurate or harmful medical and emotional advice.

But other critics are part of a large and influential online community called rationalists, or effective altruists, who believe AI could eventually destroy humanity. This state of mind is reflected in the letter.

Please share your thoughts and feedback on our On Tech: AI series by answering this short survey.

Your homework

We can speculate where AI will go in the distant future, but we can also ask the chatbots themselves. For your final assignment, treat ChatGPT, Bing, or Bard like an enthusiastic young candidate and ask where he sees himself in 10 years. As always, share the answers in the comments.

Quiz

Question 1 of 3

What functionality has OpenAI demonstrated with GPT-4 that is not yet publicly available?

Start the quiz by choosing your answer.

Glossary

Alignment: Attempts by AI researchers and ethicists to ensure that artificial intelligences act in accordance with the values ​​and goals of the people who create them.

Multimodal systems: ChatGPT-like AIs that can also process images, video, audio, and other non-text input and output.

General artificial intelligence: An artificial intelligence that matches the human intellect and can do anything the human brain can do.

Click here for more glossary terms.

Farewell

Kevin here. Thank you for spending the last five days with us. It was a pleasure to see your comments and your creativity. (I especially liked the commenter who used ChatGPT to write a cover letter for my job.)

The subject of AI is so vast and fast-paced that even five newsletters are not enough to cover everything. If you want to dive deeper, you can check out my book, “Futureproof,” and Cade’s book, “Genius Makers,” both of which detail the topics we covered this week.

Cade here: My favorite comment came from someone who asked ChatGPT to plan a route through their state’s trails. The bot ended up suggesting a trail that didn’t exist as a way to hike between two other trails that do exist.

This little snafu offers a window into the power and limitations of today’s chatbots and other AI systems. They have learned a lot from what is published on the Internet and can use what they have learned in a remarkable way, but there is always the risk that they insert plausible but false information. Go ahead! Chat with these bots! But trust your own judgement, too!

Please take this short survey to share your thoughts and feedback on this limited edition newsletter.