Now that we've come out the other side of the holidays, 'tis the season for predictions about what awaits us in the coming year, and the MIT Technology Review just served up its list of the 10 biggest technology breakthroughs that it expects to occur in 2025.
One of the AI applications cited could accelerate the efficiencies insurance companies are finding with generative AI, and another could have a profound effect on how insurance companies market themselves and generate leads. Still others could make 2025 the year of the robotaxi and possibly get you closer to having that robot valet you've always wanted. Other technologies on the list would have far smaller impacts for the insurance industry but would create challenges and opportunities by changing the risks in major industries, such as steel and ranching, while slowing the rate at which the world is heating up.
The Review isn't always right -- and I have reservations about a couple of their projections -- but the publication is always thought-provoking.
Let's have a look at what they expect for this year.
At the top of the list is an observatory that will come online in Chile this year and help scientists unravel mysteries such as dark matter, but, as interested as I'll be in what's learned, the observatory won't affect insurance, so let's move on to No. 2.
No. 2 is generative AI, in particular its ability to synthesize information for users of search engines, rather than take the traditional approach of serving up links. The article says: "Google’s introduction of AI Overviews, powered by its Gemini language model, will alter how billions of people search the internet. And generative search may be the first step toward an AI agent that handles any question you have or task you need done."
The shift, which I agree will start happening in a big way this year, will disrupt marketing and sales-generation efforts that are designed to spot people asking certain questions and steer them to a corporate website for answers and, eventually, a purchase. The shift will also cause great strains for media companies, already under siege, because the search engines will now simply answer queries rather than steer users to the media outlets to read articles with the underlying information. What the change does for advertising, in particular, and journalism, in general, is hard to project, but the changes will be earth-shaking -- and, I expect, not good for the craft I've practiced for decades now.
I'm less sanguine about the prospects for so-called agentic AI in the short term. While insurance companies will get much better this year at using generative AI to more efficiently gather, process and generate documents for claims agents, underwriters and agents and brokers this year, I just don't think the technology is mature enough yet to actually let an AI act independently on behalf of a human worker. Even if the technology is ready, the trust isn't there. We all know the horror stories about AI hallucinations, and the consequences are far greater if my AI can act as me rather than just presenting information for me to review. Agentic AI will happen, but I'd put it on my calendar for 2026 or 2027.
No. 3 is small language models. That term is designed to contrast with the large language models (LLMs) that ChatGPT, Gemini and others have developed over the past two years by consuming every bit of information they can find anywhere about anything. I've been waiting for these small models almost since I first heard about LLMs, and I think the Review is right that they should blossom this year.
Small language models benefit from what the developers of LLMs have learned but can be focused just on the insurance industry, or a segment, even on a particular company's client and knowledge base. As a result, the small models can be more precise and accurate as they serve up automated responses to clients or as they address questions on data or policy by claims agents, underwriters and others.
Because the models are small, they require much less computation, so they're faster and use far less electricity. I expect a major impact on insurers, starting this year.
No. 4 is "cattle-burping remedies." You read that right. Cows burp methane, a far more pernicious greenhouse gas than carbon dioxide. Cows are the biggest single source of livestock emissions, which together account for as much as 20% of the world's total climate pollution. An old friend of mine is a co-founder of a company, Blue Ocean Barns, that I believe is the largest producer in the U.S. of the food additive that can reduce methane emissions 30% in dairy cattle and much more in beef cattle, so I've been reading about her work for some time and agree that the additives are both proven and ready to scale. I hope so.
These remedies will have little effect on the insurance industry but will change some of the risks in the massive dairy farming and ranching industries.
No. 5 gets us back to AI, in this case in driverless cars. The Review says 2025 will be a breakout year for robotaxis, and I agree. While the rollout has taken longer than many of us expected a few years ago, Waymo is now operating a driverless-taxi service in San Francisco, Los Angeles and Phoenix and plans to expand into Austin and Atlanta later this year. Others, including Amazon, are beginning to operate robotaxis in test markets. Yes, General Motors stopped funding the robotaxi efforts of its Cruise subsidiary last year, and I still see no reason to take Tesla seriously despite all the robotaxi hype built into the stock price, but I think we've passed a tipping point.
The services will still be so limited in 2025 that they shouldn't have any noticeable effect on the auto insurance market, but the change will only pick up speed from here.
No. 6 is cleaner jet fuels. Like the food additives for cows, making jet fuel from industrial waste or carbon dioxide pulled from the air involves some really cool stuff and could be good for the environment but will have little effect on insurance other than to change some considerations about airlines and those who make their fuel.
No. 7? AI again, this time in the form of fast-learning robots, using many of the same techniques that LLMs do for generative AI. The Review says the acceleration of learning already means robots are washing dishes in industrial settings and says they soon could help out at home, too.
I'm skeptical. I've been waiting for my robot valet ever since watching "The Jetsons" as a kid, and I've read far too many predictions like this one to think we've finally arrived. I have no doubt that robots are making stunning progress in settings such as Amazon warehouses, but you'd have to wash an awful lot of dishes at home to justify spending tens or hundreds of thousands of dollars on a robot. And what would the robot do when it isn't washing dishes? Can it clear the table and put the dishes away after washing them, while staying out of my way? Can it also vacuum, clean up after the kids, walk the dog...?
We'll get there, but maybe in 2035, not 2025. In any case, I don't see major implications for insurers any time soon.
No. 8, long-acting HIV prevention meds, and No. 10, stem-cell therapies that can treat epilepsy, type 1 diabetes and more, would be wonderful, but even the Review doesn't expect major results for years. No. 9, "green" steel, fits in with cattle-burping remedies and cleaner jet fuels in my book. The steel industry generates about 8% of the world's greenhouse gases, so cutting way back on its production of carbon dioxide would be a huge step in the right direction, but the effect on insurers would be minimal.
Here's hoping the insurance industry will, as usual, do its part in quantifying, mitigating and transferring risk so these 10 breakthroughs, and many more, can flourish this year.
Happy New Year.
Paul
P.S. In case you're curious, here is the list that the MIT Technology Review produced at the beginning of 2024 on expected breakthroughs last year. I'd say they nailed it on "AI for everything," super-efficient solar cells, weight loss drugs, Twitter killers, and the first gene-editing treatment. They were mostly right on enhanced geothermal systems, heat pumps, chiplets and exascale computers. They only missed on the Apple Vision Pro (which I told you back in February would be a flop, based on the same sort of experience and analysis that convinces me that the Tesla robotaxi and robot valets are nowhere near ready).
P.P.S. Here are the two smartest pieces I've read thus far about the innovations being unveiled at CES: The 10 Coolest Things We’ve Seen So Far at CES 2025 and AI takes over CES 2025 — one smart gadget at a time. Nothing in either article does much for me, but, then, I'm not a gadget guy. I was given a page turner for my Kindle for Christmas and was genuinely puzzled at the concept. My finger doesn't work any more? But maybe you're into gadgets more than I am.