People are increasingly worried about being replaced by AI. While powerful, current AI tools are still very limited. As they evolve, productivity will accelerate and new jobs will be created.
A study from OpenAI and the University of Pennsylvania found that “around 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of LLMs”, LLMs being Large Language Models like OpenAI’s GPT.
Research from Goldman Sachs suggests that 300 million jobs could be affected by these tools, but they could also boost global GDP by 7% over time.
There are two important caveats here:
Firstly, current AI is not yet AGI (Artificial General Intelligence), so we must be careful not to exaggerate its abilities. It’s not the superintelligent artificial brain that some people believe it to be. It’s ‘just’ a very powerful type of software that was trained on a huge chunk of all publicly accessible knowledge.
Secondly, while some jobs are replaced by AI, many are positively affected by it, and even more will be created because of it.
Tools disrupt work, but they can’t fully replace humans
Large Language Models (LLM) are essentially probabilistic prediction software. They’re trained to identify patterns in huge amounts of data, enabling them to reply to questions and follow instructions. Whether the replies are correct or not is a function of probability, and can be judged by humans and separate models. To improve a model requires a new round of training with updated data and user feedback.
The result of all of this is a program that you can talk to, and it will predict the most probable answer to your question. This is the need-to-know gist of it, though obviously there’s much more to how LLMs work. After all, GPT4, the strongest LLM yet, is “the most complex software humanity has likely produced” as described by an applied AI expert from OpenAI.
This program is powerful and fascinating, but limited. It can’t be a world-class tech journalist yet, as the blowback against CNET’s AI content showed. It lies too much to be useful to lawyers, as one found out after ChatGPT added fake citations and opinions to his legal brief. It’s not smart enough to replace human consultants at a helpline for eating disorder sufferers.
Building it requires vast amounts of data and computing power. Despite all the refinement and optimization, because of the probabilistic nature of these models, it hallucinates things that aren’t true. Especially in use cases that require precision and deterministic logic, like math and physics, or citation of true sources. It can’t deal with novelty, empathize with the human experience, or interact with the physical world the way we can.
Just like translation software didn’t remove the need for human translators, this new software won’t remove the need for human knowledge workers. But it will augment us, make us much faster, and allow us to focus on what humans do best – making new tools, critical thinking, imagining bold ideas, making tough decisions, improving the world around us, and caring for other humans.
History proves that the economy can accommodate any increase in productivity, no matter how large. We will use the huge efficiency boost from LLMs (and the technologies that will follow) to create new types of products, services, and whole industries that will generate jobs for humans.
The invention of electricity made life easier with basic comforts like artificial lighting, air conditioning, and useful home appliances. These were all luxuries in the past, but are now standard. They have become table stakes in our lives, and generative AI is poised to join them. In a few years, technology that isn’t AI-enhanced will seem outdated and useless. Us, humans, are very good at expecting a greater quality of life whenever there is such a possibility.
The truly amazing use cases for AI tools
After ChatGPT came out, many people predicted that it would eat Google’s search business. It’s been a while, and now we have Bing chat search, Google is close to releasing a genAI-powered search experience, and there are third-party AI search engines like Perplexity. There was a bit of hype but it fizzled out because out of the many uses for LLMs, disrupting search is an underwhelming one.
This shows that nobody can really predict the consequences of a revolutionary technology. Which is comforting when you consider the doomsayers that have been coming out against further AI development.
There was the infamous open letter to pause AI development, signed by 1,000+ influencers including Steve Wozniak and Elon Musk. It was criticized for vagueness and looking like an effort to stifle OpenAI’s growth potential in order to let others catch up. More recently, a non-profit called the Center for AI Safety released a statement warning about the many dangers of AI.
AI should both provide the most individual economic empowerment of any technology ever and also make it so that no one has to work who doesn’t want to— Sam Altman (@sama) December 16, 2022
It’s good to discuss the negative side of things, but the companies that build these systems seem to be on top of the situation. For probably the first time in history when it comes to innovative technologies, they are pre-emptively asking governments to quickly implement AI regulations. They’re helping to outline policies and best practices for AI security.
Meanwhile, legions of smart people around the world are hard at work trying to maximize the usefulness of AI in important and exciting areas, such as:
- Combatting the shortage of healthcare workers with bots that can diagnose diseases with great accuracy.
- Improving the quality of education for everyone with AI tutors like the one Khan Academy is working on.
- Helping scientists and engineers discover new solutions faster, for example researchers that used ChatGPT to improve antibody therapies.
- Optimizing business management and execution, and providing much-needed help for small businesses.
- Making government services available to more people, like it’s already happening in India.
We’re faced with climate change, geopolitical volatility, societal divisions, and economic stagnation. We need more productivity and more brain power to address these challenges. Here, AI tools will help us the most.
To accelerate, or not to accelerate?
Bad actors will use AI tools for bad purposes, but such is the case with every technological breakthrough. Instead of pausing development, we should focus on education and prevention.
Even if the current generation of AI tools ends up having an underwhelming impact on the world, the next generation will make up for it. So, for business and government leaders, as well as for any knowledge worker, there is no question of “should I start using AI?”.
The only question is “why aren’t you using it yet?!”