‘Mind-blowing’ new AI can do ‘almost anything’
Should we be worried by AI? A groundbreaking program called GPT-3 can learn almost anything without human help. The possibilities are limitless, but some are saying so are the risks.
Did a human or a computer write this story? It just got much harder to tell.
A new artificial intelligence program called GPT-3 is creating a wave of excitement among scientists with its astonishing ability to learn and use language. Programmed with 175 billion rules and fed half a trillion words, GPT-3 is by far the most powerful language model ever built.
“Playing with GPT-3 feels like seeing the future,” says tech developer Arram Sabeti, who got the AI to write stories, articles, songs, and poetry. Other users described as “mind-blowing” its ability to code websites, compose music, and even diagnose diseases.
This ability to master all kinds of specific tasks without additional human input is a technological breakthrough with enormous implications.
But not everyone is enthusiastic. Entrepreneur Elon Musk has described AI technology as “summoning a demon” – even though he is one of the main funders of the company that designed GPT-3. He fears that we are approaching the technological singularity when computers will develop out of control and become more intelligent than humans.
We already live in a world run by AI. Many of us happily let it predict the words we type, choose our next song, control the temperature of our home, and even fly us 30,000 feet above the Earth. Because AIs don’t get tired or make human errors, they are also making driving much safer.
Others have raised warning voices. The rise of robots threatens many people’s jobs. And there is something particularly creepy about machines that appear to think and behave like humans.
The possibility that a machine may become conscious is a fear deeply rooted in science fiction, from Frankenstein to The Terminator.
However, GPT-3 is not conscious. It lacks many of the features of human intelligence, including common sense and narrative reasoning. Ask it a nonsensical question like, how many eyes does my foot have? It guesses two. GPT-3 has read half a trillion words, but it does not have an abstract understanding of what feet and eyes are.
Nor can it tell a complex story. We can hold a complex narrative in our mind over hundreds of pages. GPT-3 writes word-by-word, making grammatically correct sentences that often take surreal tangents and can quickly lose the plot. Writers and artists should not fear losing their jobs any time soon.
AI programs may not think like us, but that doesn’t stop them being dangerous in other ways. In 1950, Alan Turing designed a test for artificial intelligence to see if a computer could be mistaken for a human. By 2018, chatbots were getting close to passing the test. GPT-3 performs even better.
Critics say this ability to impersonate humans is the biggest risk. GPT-3 can read everything you have written and then pretend to be you, and convince other people that it is you. People who want to spread disinformation and fake news will find a powerful new tool.
So, should we be worried?
Deus ex machina
No. Humans have always been suspicious of new technology and we should embrace the wonders of artificial intelligence. Instead of worrying about it replacing us, we can focus on how AI will allow us to be more creative. Not only will we spend less time on routine tasks, but AI will give us easy access to a limitless source of information.
Yes. For the first time in human history, we are developing a tool that can learn to improve itself. The developers of AI admit they don’t know the full capabilities of their creation, but their goal is to develop a computer that surpasses human intelligence. This should frighten us because once it is achieved, there is no going back.
- Are super-intelligent computers exciting or terrifying?
- Will AI ever be more intelligent than humans?
- Draw your very own personal super-robot.
- Write a story from the point of view of an intelligent robot that has just become self-aware.
Some People Say...
“The development of full artificial intelligence could spell the end of the human race.”Stephen Hawking (1942-2018), English physicist
What do you think?
Q & A
- What do we know?
- It is generally agreed that the development of artificial intelligence is speeding up. The first AI was built in 1956 and was a simple computer program for proving mathematical theorems. Last year, GPT-3’s predecessor operated with 1.5 billion rules and was described as “too dangerous to release” by its developers. GPT-3 has 77 billion rules and is hailed as “groundbreaking”. Some suggest that, at this rate, AI language models will be as complex as the human brain in a few years.
- What do we not know?
- One main area of debate is around how concerned we should be about these developments. Optimists believe that artificial intelligence will never replace nor fully replicate the complexity of the human mind. Or if it does, we will be able to control that intelligence and use it for our benefit. Pessimists argue we need to control and regulate this new science now because, by the time a super-intelligent program has been developed, it may well be too late.
- Artificial intelligence
- AI refers to any machine, robot, or software that performs complex tasks independently, without requiring instructions from humans.
- This is the third generation of the Generative Pre-training Transformer. It uses machine learning to develop new skills and does not need to be programmed like traditional software.
- Technological singularity
- As AI learns and self-improves without human intervention, its progress will accelerate beyond our control until it becomes indistinguishable from humans, and computers and humans become one (singular) race. Many computer scientists think this point may be reached within our lifetime.
- Driving much safer
- Most modern cars have automated systems, like parking assistance, to reduce driver errors. The first fully self-driving cars are expected to be ready in the next couple of years.
- People’s jobs
- In 2013, a study from Oxford University predicted that 47% of US jobs could be lost due to automation. A popular website, Will Robots Take My Job, gives the probability of your job being safe from robots. (Teaching jobs are classed as “totally safe”.)
- Frankenstein to The Terminator
- Mary Shelley’s 1818 novel dealt with the idea of humanity playing God, inventing life, and losing control. In The Terminator franchise, artificial intelligence becomes self-aware on 29 August 1997 and starts a war with humanity.
- Going off into a completely new subject.
- Alan Turing
- The English mathematician is considered the father of AI. He played a major role in cracking the German Enigma machine in WWII.
- Software applications used to conduct on-line conversations via text or text-to-speech, instead of providing direct contact with a live human.
- Artificial intelligence is already used to produce “deep fakes”, visual and audio content that impersonates someone in order to deceive and spread false information.