Why is AI is hitting the headlines in 2023, is your job at threat and who will the winners and losers be?

Since its demo was publicly released on November 30 last year, OpenAI’s ChatGPT-3 large language model (LLM) AI tool has captured the imagination. Testers have been hugely impressed by the tool’s ability to usefully and (usually) correctly answer queries on anything and everything.

Ask it how to code a function in a given programming language and it will spit out a line of code that will, in most cases, work. It was famously able to pass Google’s coding interview for the tech giant’s ‘Level 3’ engineers, which earn a $183k salary. The tool’s API powers Microsoft-owned GitHub’s CoPilot, which turns natural language prompts into coding suggestions across dozens of languages.

Marketers are using it for tasks such as suggesting a structure for an article on a particular topic, including headings and sub-headings. Some are even getting it to write each sub-section, churning out entirely AI-generated content that can pass for a human writer and researcher’s work.

Students, of course, quickly cottoned on to its potential as a home work “assistant”. If they were, for example, given a homework assignment to explain ChatGPT, the tool itself could help them submit a comprehensive effort that involved minimum work.

What can ChatGPT-3 and other LLM like Google’s Apprentice Bard do?

ChatGPT and other LLM can answer questions based on the content they have been trained on, in ChatGPT’s case

The buzz that followed OpenAI’s release of its generative AI tool, and news that its backer Microsoft is integrating it into products such as its Office productivity suite (the days of struggling for the right Excel formula appear numbered) and even Bing search engine, has been growing rather than fading away.

ChatGPT has been the catalyst for more attention to be given to other AI tools that also have the potential to change the way we live and work in coming years. OpenAI, a “capped profit” company committed to reinvesting all future revenues beyond a fixed threshold back into its not-for-profit AI R&D foundation, has other tools including the generative AI image generator DALL·E.

Rumours that Google were spooked by the possible implications of ChatGPT to its domination of the search engine market appear to have had some foundation in reality. This week Google unveiled its own beta LLM tool, Google Bard, which works in roughly the same way. It also has the same limitations, confidently answering one question wrongly while being publicly demoed.

Other big tech companies have been ploughing billions into their own LLM models and other forms of AI. OpenAI’s ChatGPT is far from the only show in town, even if its quickly become synonymous with the crescendo of hype around AI in early 2023.

The quality of ChatGPT’s output is tightly connected to the quality of the prompts the tool is given. For example, this follow-up prompt results in a useful expansion on how the technology works:

And what was the language-based data the ChatGTP-3 LLM was trained on?

And a further prompt results in greater detail.

What are the limitations of LLM?

By their very nature, AI tools like ChatGTP and Bard can never come up with anything genuinely innovative. They simply regurgitate the material they have been trained on, though are skilled at putting elements of it together in a way that sounds eerily like a human in response to human questions. But because the output is based on predicting what should come next based on data patterns, rather than putting pieces of information together by “thinking” like humans do, it can, and regularly does, make mistakes. And sound extremely confident while doing so.

By their very nature, LLMs like Chat GTP cannot be better than humans at anything except tasks which are precisely, and not more than, accurately and efficiently representing what a lot of people have said.

Do LLM like ChatGPT-3 and Bard threaten jobs?

While the general public have been wowed by ChatGPT’s capabilities to the extent fears have been aired that the technology could replace swathes of white collar jobs, that is unlikely to happen. It’s important to note that the technology used by ChatGPT is neither unique nor even especially new.

Yann LeCun, Meta Platforms’ chief AI scientist recently told a press conference held via Zoom:

“It’s nothing revolutionary, although that’s the way it’s perceived in the public. “It’s just that, you know, it’s well put together, it’s nicely done.”

He continued:

“OpenAI is not particularly an advance compared to the other labs, at all. It’s not only just Google and Meta, but there are half a dozen startups that basically have very similar technology to it. I don’t want to say it’s not rocket science, but it’s really shared, there’s no secret behind it, if you will.”

The fact that the AI tools we’re starting to become aware of as a general public are based on technology many companies have access to and are able to work with means they will become commoditised as they evolve. They will change the way we work and do certain things over coming years and certain functions traditionally requiring human labour may be executed by AI tools with minimal human input.

However, the fact that almost everyone will have the same access to these tools and capabilities firstly means nobody will have a real edge. They’ll become a standard part of work processes in the way productivity tools like Excel or alternatives like Google Sheets have.

Individual jobs may be made redundant by AI. But in the same way the introduction of computers into offices and automation in factories and on farms didn’t lead to mass global unemployment, tools like ChatGPT and future iterations and other AI models are unlikely to either. They will simply increase the amount of output, or productivity, of individual workers using them.

Who will be the commercial winners and losers in the AI race?

Big Tech’s labs and start-up challengers, the most successful of which are likely to be promptly bought up by Big Tech, are in a race to develop AI capabilities that will give them a genuine edge on rivals. There’s not currently any clear frontrunner. However, as The Economist notes:

“All recent breakthroughs in big AI globally have come from giant companies, because they have the computing power, and because this is a rare area where results of basic research can be rapidly incorporated into products.”

“Amazon, whose AI powers its Alexa voice assistant, and Meta, which made waves recently when one of its models beat human players at “Diplomacy”, a strategy board game, respectively produce two-thirds and four-fifths as much ai research as Stanford University, a bastion of computer-science eggheads. Alphabet and Microsoft churn out considerably more, and that is not including DeepMind, Google Research’s sister lab which the parent company acquired in 2014, and the Microsoft-affiliated OpenAI”.

Ultimately, the winners in the AI race will not be those who develop the most innovative and ground-breaking new developments in the space from a scientific point of view. There is no profitability in AI for AI’s sake, no matter how impressive its output. It will boil down to who comes out with most commercially valuable products that are either made possible by AI or are powered by it.

Drugs discovery is one sector AI is having a huge impact on but goes relatively under the radar because the public don’t see it. Computer vision, a sub-discipline of AI the Chinese are thought to have taken a lead in, is another area whose integration into commercial products is still nascent but has huge potential. Some of it, like the potential for mass state surveillance capabilities, concerning.

Big Tech in both the USA and China will almost certainly lead the way when it comes to the commercial exploitation of AI. However, the capital flowing into generative-AI startups, which last year collectively raised $2.7bn in 110 deals, suggests that venture capitalists are betting that not all the value will be captured by big tech.

Tools like ChatGPT-3 are impressive and they and others powered by different kinds of AI models, or combinations of models, will change the way we do many things going forward. But the pace of change is unlikely to be as swift as the hype around AI since the end of last year might suggest. Professional roles will be changed by the introduction of new AI powered tools but the large majority are more likely to evolve than be lost.

AI represents an exciting technological leap forward for humanity but it’s one that has been brewing for a long time now and will make itself felt in our lives incrementally over the next decade and beyond. Social and traditional media buzz around ChatGPT has brought AI into the public discourse and increased awareness. But there hasn’t been any large leap forward made by OpenAI or any of the other leading AI labs. It just looks that way.