- Investment is important but not enough as the field of AI needs more innovation and experimentation, Frey tells Anadolu- Current data-intensive approach to AI means ‘only a few companies can actually afford to push the frontiers,’ says FreyLONDON
Artificial intelligence technologies could well prove to be the “most transformative thing” humans have ever invented, but their impact on production processes and the global economy so far has been “relatively disappointing,” with a raft of factors hindering substantive progress, according to a leading expert in the field.
Over centuries, humans have used their own intelligence to create a lot of new technologies that gave us longer and better lives, but that progress has come with costs such as climate change and destructive weapons, said Carl Benedikt Frey, professor of AI & Work at the University of Oxford.
“In that regard, we’re not in uncharted territory. What may be different is that artificial intelligence, as a general-purpose technology, is likely to transform just about every aspect of life,” he told Anadolu.
“The potential applications are so many that it’s very hard to know exactly what the risks are going forward, so it very much depends on what we use it for.”
Comparing the development of AI to past milestones, he said it has the potential to be on par with the agricultural and industrial revolutions.
“But, I think, in terms of its impact on productivity and the economy, it’s been relatively disappointing so far. It’s nowhere near what we saw with electricity or the internal combustion engine. We haven’t seen anything like the transformation we saw with the automobile, where the bulk of the population all of a sudden had access to personalized transportation,” said Frey.
“As a consequence, it’s very much a question of where this technology goes over the next couple of years. It’s certainly plausible that it will be the most transformative thing that we have invented … but to be really on par with some of these previous technological milestones, we need more progress.”
He stressed the need for “more innovation in the field … and AI that is more robust,” not just limited to memorizing “data and information that humans have produced.”
The requirement now is AI that is “actually capable of adjusting to new circumstances and is resilient, as well as capable of planning and learning from smaller datasets similar to humans,” he added.
‘What can we do with AI that we couldn’t previously?’
On the big question of how AI will impact human workforces, Frey asserted that the answer is intrinsically linked to how the technology is utilized in any specific sector.
“If we use AI for automation, the impacts will be job displacement, pressure on wages, falling labor share in national income,” he said, adding that this would be very similar to what was seen during the first industrial revolution.
The focus then was “on replacing our workforce and existing activities through, at the time, mechanized factories, which gradually took over production from the domestic systems, where people worked in their homes,” he explained.
“As production mechanized, the jobs of craftsmen vanished,” he said.
“That’s quite different from what we saw in the middle part of the 20th century, where the automobile and electrical industries created vast new operations that absorbed a lot of people.”
That happened because of “technologies which created new products, new tasks, new industries for workers,” he said.
“If we look at existing applications of AI, it looks to be more of a replacement technology. Nearly all applications I can think of focus on doing something that we’re already doing a bit more productively,” he said.
“I think the question we need to ask ourselves going forward is what can we do with artificial intelligence that we couldn’t previously do, or what new pathways to discovery does it open ... rather than asking how can I be a bit more productive in what I’m already doing.”
Another change that Frey views as necessary is striking a balance between investment and experimentation, stressing that the latter is vital for substantive progress.
“The Soviet Union invested a huge amount … A lot of large corporations that no longer exist invested a lot, but invested in the wrong things. So, investment by itself is not ideal and does not lead automatically to innovation and technological progress,” he said.
The field of AI needs exploration right now, which means “entrepreneurs taking different types of bets, and not just following one technological pathway,” he said.
“I think a key concern right now is that current approach to AI is so data- and computer-intensive that only a few companies can actually afford to push the frontiers in that particular space,” said Frey.
What we need is to “incentivize other players to come in and experiment with less data-intensive models,” he said.
“So, hopefully we will have more of that experimentation happening going forward. Investment is surely important … but it’s not enough.”
