Despite all the doom and gloom, AI could have a huge positive impact on medicine and the economy.

It’s Science Fiction with Stuart Ritchie, a newsletter exclusively for i-subscribers. If you’d like to receive it straight to your inbox every week, sign up here.

This is one of the most important weeks for AI. Not only is there an AI Security Summit taking place at Bletchley Park (in fact, the UK government has invited all the leading AI experts to come and discuss how we will control AI technology in the future), but also that US President Joe Biden issued an executive order on how the United States will regulate the development of AI in the future.

This is all the result of months (and sometimes years) of concern about the potential dangers of AI. While they have long presented challenges and opportunities, 2023 has truly shown how advanced AI models have become.

Anyone who’s used a chatbot like Open AI’s ChatGPT or Anthropic’s Claude, or an image generator like DALL-E will be impressed with its capabilities (and if you haven’t tried it yet, why the hell not? !).

Think of it this way: Almost any task that the human brain can perform, AI can likely do faster and more efficiently. Unfortunately, the human brain can do many bad things: lie and cheat; obsessive control over other people; have prejudices against certain groups; Making weapons. Artificial intelligence models are already making these things much easier. In the medium term, increasing model efficiency will put many human jobs at risk. Things could get very bleak in the long term: I’ve written before about the existential risks of AI.

Many people argue that these negative aspects of AI mean that AI companies should simply stop working on these models. Why develop something that can have such terrible flaws?

Let’s take this question seriously and look at some possible answers. What are the benefits of developing advanced artificial intelligence models?

Economic productivity

We hear it endlessly: we are experiencing a productivity slump. The UK economy is not growing nearly as fast as we would like, and the plans of various politicians over the years have failed to accelerate this growth. Can AI be the game changer we need?

More and more companies are already using AI to improve their productivity: tasks such as calendar management, transcribing meetings, summarizing documents, and more. translate between different languages all of which have suddenly become much simpler and can largely be transferred to artificial intelligence systems, freeing people up to do something more creative.

AI is also useful for higher-level tasks: patterns in data sets (such as where and when different products are sold) that human observers might never notice can be detected by AI models and help companies improve their performance. Advertise and increase sales.

Governments will also be able to use AI to improve the effectiveness of their policies by analyzing data sets collected in different parts of the country and from different institutions to improve service delivery. In the UK, artificial intelligence could help connect the country’s vaunted NHS services and prevent situations like those described this week by BBC correspondent Rory Cellan-Jones, who said communications between NHS staff in different hospitals and even the various departments are “grossly inadequate.”

These are just a few ideas. But just as computers, email, and the Internet have allowed us to become far more efficient than in the days of telegrams and faxes (and opened up entirely new types of businesses that didn’t exist before), it’s easy to see how AI works. If used correctly, it can make a huge contribution to the economy and lift us out of a long period of stagnation.

The science

Scientists have been using some components of AI, such as deep learning, for years to improve their research and get more from their data sets. But some recent events are really pushing the boundaries. For example, this week Google DeepMind announced some exciting new developments in its AlphaFold model, which attempts to discover the structure of proteins and other molecules.

Knowing the structure of molecules is one of the great open questions in biology, and it is extremely important: if you know the exact structure of a molecule, you can design a drug that fits it like a key fits a lock. This means that your drug can have a much stronger effect on everything that protein does, and therefore on the disease in which it is involved. An example is diseases like Covid: using these AI tools can help us develop antibodies that neutralize the effects of the virus. and don’t stop working as soon as a new option appears.

More generally, in many cases scientists have a lot of data and do not have the intelligence to analyze it. This is no insult to scientists: no human can hope to fully understand how the human genome works or what happens in our brains when we perform complex tasks.

This is where AI comes into play. The pattern recognition capabilities of AI models can be applied to huge sets of scientific data and, if the data is detailed enough, can give us completely unexpected information about all sorts of important biological and chemical processes. We could suddenly make progress in the fight against diseases such as dementia and some types of cancer, where science is usually stumped.

Fighting crime

Yes, AI has great potential to cause fraud, identity theft, and other forms of financial crime. Criminals are already raving about the possibilities that AI models can offer them if they can be hacked using the ethics built into them by their manufacturers. When it comes to fraud, we are likely to see an arms race between AI for evil and AI for good, just like we have seen in the fight against spam and spam filters over the past few decades.

But other forms of crime can also be combated using AI. Systems such as the National Police Computer have already revolutionized the way the justice system uses data. I recently read about the horribly botched Yorkshire Ripper investigation of the 1970s and 1980s, in which the police interviewed the perpetrator nine times and stored his information on various account cards in their hacking database, but never linked the information together (he was eventually arrested). Accident).

Today, with the availability of computer technology, such a catastrophe would be extremely unlikely. But AI goes even further: identifying even more detailed patterns of violations helps prevent crimes and catch more criminals. AI can also be applied to forensics: we will be able to model aspects such as bullet trajectories and blood spatter to understand in more detail how a crime occurred.

A big part of fighting crime is identifying patterns. And, as we saw above, AI is already much better than humans.

Entertainment

If you’re an avid Spotify user like me, you’ve noticed that artificial intelligence is also being used in the entertainment industry. Spotify’s new AI DJ leaves a lot to be desired at the moment (no, I don’t have to listen to a 16-minute piece of classical music with a 3-minute pop-rock song – I don’t). But the algorithms have an almost uncanny level of precision in reproducing music I like that I’ve never heard of before, and they’ve only gotten better over time.

We also saw this in Peter Jackson’s recent film. Come back The Beatles suggest that artificial intelligence techniques can be used to improve old films, moving from drab and boring images to beautiful, colorful details.

Not only can AI help us enjoy things that already exist, but it can also be used to create entertainment. In the future, we will see video games where non-player characters can have long, detailed, and unique conversations with each player using chatbots tailored to the character’s specific personality.

Of course, one of the reasons for the recent writers’ strike in Hollywood was the fear that AI could replace human writers in films and TV shows. And many authors and artists are outraged that AI models are trained on data from their work without consultation or payment (in my opinion, this is a lot of what economists call “rent seeking”; people want to get paid). for actual failure to provide a useful service).

As with any new mainstream technology, there will be a lot of back and forth (the music industry has only recently, after years of fighting piracy, really embraced the streaming model – and there are still protests about how she works). But it would be foolish to argue that AI won’t bring great benefits to so many areas of the entertainment industry.

Look on the bright side

Don’t get me wrong: none of the above benefits mean we should be distracted when it comes to AI safety. I’m very glad that we held the AI ​​Safety Summit and pushed all AI companies to take various measures to ensure that their models remain under our control and fit for human purposes.

But amidst all this gloom and doom, we must remember that the real reason we develop these models in the first place is to improve the human condition. Despite the shortcomings, the sheer power of AI means we have a very good chance of achieving just that.

Other things I wrote this week

When I recently started weightlifting at the gym, I wondered which of the many supplements, pills and powders available on the market could actually help improve your performance. I have reviewed the evidence of some of the most prominent of these.

Science Link of the Week

I think this is a very sad article about how the quality of Japanese research has declined over time. Is this due to the aging population and low immigration in the country? Anything else? Whatever it is, it is a cautionary tale that other countries want to avoid.

It’s Science Fiction with Stuart Ritchie, a newsletter exclusively for i-subscribers. If you’d like to receive it straight to your inbox every week, sign up here.