Want to search for posts across your favourite author? Just click on their name to reach their '100 most recent posts' page, and then use the search box located beside their name.
- Manlobbi
Halls of Shrewd'm / US Policy❤
No. of Recommendations: 10
The Things AI Won't Change
"
In a world facing rapid change from AI, it's the things that won't change that you want to build on and invest in.
That's the classic perspective from Jeff Bezos. He famously told The Harvard Business Review in 2007:
It helps to base your strategy on things that won't change. When I'm talking with people outside the company, there's a question that comes up very commonly: 'What's going to change in the next five to ten years?' But I very rarely get asked 'What's not going to change in the next five to ten years?' At Amazon we're always trying to figure that out, because you can really spin up flywheels around those things. All the energy you invest in them today will still be paying you dividends ten years from now. Whereas if you base your strategy first and foremost on more transitory things'who your competitors are, what kind of technologies are available, and so on'those things are going to change so rapidly that you're going to have to change your strategy very rapidly, too.
Bezos believed people would always want great retail selection, low prices, and fast delivery regardless of new technology like the Internet. The Internet was a conduit to meet those persistent demands. He was right, and people still want those same things today almost 30 years after the founding of Amazon."https://deepwatermgmt.com/the-things-ai-wont-chang...One looks at Berkshire's businesses and stock portfolio holdings and sees that most of BRK's investments (insurance, food, energy, real estate) are largely "things that won't change."
No. of Recommendations: 2
One looks at Berkshire's businesses and stock portfolio holdings and sees that most of BRK's investments (insurance, food, energy, real estate) are largely "things that won't change.How can AI NOT change insurance? Personalized premiums, claims fraud, risk analysis, everything that used to depend on actuaries' skills and experience.
This is assuming AI is not just hype and can justifiably (XAI -
https://en.wikipedia.org/wiki/Explainable_artifici...) predict better than the experts, so as to avoid or win lawsuits.
No. of Recommendations: 1
"How can AI NOT change insurance?"
It won't change the need for insurance.
No. of Recommendations: 10
How can AI NOT change insurance?
Replace "AI" with "software" or "data mining".
The core of being an insurer is assessing risk so it can be priced properly.
Data mining is nothing new in this context. There are potentially lots of individual observations, but only so many predictive dimensions/variables. I doubt having lots of hidden network levels can offer much accuracy beyond what any good programming team could manage already.
If all insurance firms use the available data to about the same level of accuracy, it doesn't change the competitive landscape at all.
If some companies can use the available data to assess risk better than the others, those companies will do very much better.
So the most important point is that you definitely want to be invested in a company among those that use it best.
Doing the data mining well is a very big deal, but using a specific shiny new buzzword for it seems less of a big deal.
The newer deep models might have more impact in honing the advertising copy than in assessing underwriting risk.
So I agree it will change the insurance business : )
You might be right about the potential for improving fraud detection, though in some ways that's just the flip side of risk assessment.
Better than the UK post office has managed it, one would hope.
Jim
(doing large scale data mining using neural networks, among other things, in/since the 1990s)
No. of Recommendations: 0
It won't change the need for insurance.
What human / commercial/ economic needs would change because of AI? None. Maybe some people will "need" (want) things that don't exist now. But no business as such will go away.
By this logic Berkshire should own farms and trawlers too.
AI will be bad for BRK's business if a more nimble competitor undercuts their premiums and customer service. Like Tesla is doing to Detroit with a new technology.
No. of Recommendations: 8
"In a world facing rapid change from AI, it's the things that won't change that you want to build on and invest in.
Personally I'm more concerned about investment climate impacts than investment AI impacts.
Not hurricane losses, which insurers can price incrementally, but more indirect things.
Am I invested in any businesses that are at material risk of a very bad impact? Critical dependency on facilities at flood risk or water availability? Non obvious consequences of widespread migration and conflict?
A random example: Brookfield has a lot of assets in Manhattan and London at low elevations.
Much of the Isle of Dogs, home to Canary Wharf properties, is at an elevation of 2-4m AMSL.
The small true eponymous Canary Wharf area is higher.
A sea level rise of 10m is "already baked in", much of it from thermal expansion, though it will take a very long while. I'm a short term investor, in that context.
Jim
No. of Recommendations: 0
A sea level rise of 10m is "already baked in", much of it from thermal expansion, though it will take a very long while. I'm a short term investor, in that context.
Good perspective. Current sea level rise has been very linear, in the range of 3mm/year. During the Holocene transgression (end of the last ice age) warming was extreme, and massive ice sheets covering much of North America retreated. Sea level rise averaged 11 mm/year. Without the massive ice sheets (Greenland is only huge on a Mercator map), it is difficult to come up with any scenario faster than that.
No. of Recommendations: 19
Current sea level rise has been very linear, in the range of 3mm/year. Doesn't seem like much, but it does add up.
Even if CO2 concentration stopped rising today, sea levels will still be rising 1000 years from now.
And at a glance, recent rates are pretty linear but not quite linear.
https://www.climate.gov/media/15200A big change in the rate any time soon is pretty unlikely, but even a very slow rate of acceleration can make a big difference over time, like an ion thruster.
The rate of change was about 3.34 mm/year in the last 10 years (comparing 10 year average of observations to the 10 year average from 10 years earlier)
The rate of change was about 3.02 mm/year in the prior 10 years
The rate of change was pretty linear at about 1.76 mm/year in the 70 years before that.
Slightly different rate measurement methods will give slightly different numbers, but there's definitely a speed-up.
So I doubt that the recent rate is the peak rate that will be seen.
Of course I don't have all that many more decade stretches to worry about personally, but I presume the acceleration itself will continue at least until CO2 concentrations stop rising.
And I personally doubt CO2 concentrations will ever stop rising until it is exogenously forced to happen. A shortage of affordable things to burn, or a shrinking of the global society that finds, transports, and burns them. In short, politics will fail utterly ever to bring about "net zero" by setting rules and incentives and subsidies and penalties and laws, but eventually economics (and demographics) will force it. Probably too late and in a bad way.
Jim
No. of Recommendations: 9
A big change in the rate any time soon is pretty unlikely, but even a very slow rate of acceleration can make a big difference over time, like an ion thruster.
To achieve one meter of sea level rise by 2100 the rate of rise will have to increase exponentially from present rates. That exponential curve, would have to hit a rate of 20mm/year in the period of 2080-2100, a rate twice of what occurred during the post ice age melt of the Holocene transgression. I won't be here, but I'll bet against that.
No. of Recommendations: 15
To achieve one meter of sea level rise by 2100 the rate of rise will have to increase exponentially from present rates.Who said 1m by 2100? I was thinking of 500-1000 years.
Though, as you mention it, "1m by 2100" is very near many of the current "intermediate" scenario estimates, within a VERY wide range of estimates.
e.g.,
https://www.climate.gov/media/14136As for an exponential increase in rates, the accelerating rate of rise is not in question. The second order coefficients are small, but pretty clear.
Reaching net zero soon might slow or even halt the acceleration, though not the continuing rise. Some people believe in that. Other people believe the earth is hollow.
My niece is some kind of world class expert on precisely how many people get flooded out, where, and by when.
I can post the latest numbers if you like, but it isn't exactly the sort of thing to leave you chipper. She's looking for some new, very distracting hobbies.
Jim
No. of Recommendations: 6
Though, as you mention it, "1m by 2100" is very near many of the current "intermediate" scenario estimates, within a VERY wide range of estimates.
Jim, you are very good at running some numbers to see if projections are reasonable. What rate of earnings growth and what P/E ratio is needed for Telsa and Meta to give a 10% return over the next 10 years. Often the results show the optimists to be a bit unrealistic. The same logic works with glacial geology and paleo/future sea level rise. It is very popular to be pessimistic to alarmist, and very unpopular to doubt the alarmists. Media coverage, research grants follow the eye-popping projections.
In its 2019 report, the IPCC projected 0.6 to 1.1 meters (1 to 3 feet) of global sea level rise by 2100 (or about 15 millimeters per year) if greenhouse gas emissions remain at high rates. The high end, 1.1 meters, does not pass the test for a reasonable scenario, yet it is widely quoted in media. If you ask a glacial geologist (yes many of us specialize in just that) if it is reasonable to predict sea level rise to exceed the Holocene transgression, they will say it is very unlikely. If you ask them to refute an article that Florida is going to vanish soon, they will say they'd rather not comment.
No. of Recommendations: 3
In its 2019 report, the IPCC projected 0.6 to 1.1 meters (1 to 3 feet) of global sea level rise by 2100 (or about 15 millimeters per year) if greenhouse gas emissions remain at high rates. The high end, 1.1 meters, does not pass the test for a reasonable scenario, yet it is widely quoted in media.
The forecast range is, as you say, 0.6 to 1.1 metres, under "business as usual" (the scenario I expect) by 2100.
Since that's the forecast range, I'm not sure why you'd suggest it doesn't pass the test for resonableness...it's in the range built from the best data available.
A better way to think of it is this way:
Sea levels are pretty darned certain to rise by at least a metre. It's mainly just some uncertainty of how long that will take.
From the latest IPCC report: "Over the next 2000 years, global mean sea level will rise by about 2'3 m if warming is limited to 1.5°C and 2'6 m if limited to 2°C (low confidence)."
So, we only have guesses as to when the one metre rise might happen.
Under RCP 8.5 scenario, central guess is some time around 2112. Uncertainty range is large, something like 2055-2152.
Under RCP 4.5 scenario, central guess is around 2195.
Under RCP 2.5 scenario, I think it's some time after 2300.
So, no, we're not going to see all of Florida vanish soon. You're right that way too many writers mention only the very highest and relatively unlikely estimates. But we are definitely going to see SOME of Florida disappear. Florida does somewhat justifiably get mentioned the most because of the large aggregate market value of the property that will go away, not the area.
And the chipper bit from the same report:
"The probability of low-likelihood outcomes associated with potentially very large impacts increases with higher global warming levels (high confidence). Warming substantially above the assessed very likely range for a given scenario cannot be ruled out, and there is high confidence this would lead to regional changes greater than assessed in many aspects of the climate system. Low-likelihood, high-impact outcomes could occur at regional scales even for global warming within the very likely assessed range for a given GHG emissions scenario. Global mean sea level rise above the likely range ' approaching 2 m by 2100 and in excess of 15 m by 2300 under a very high GHG emissions scenario (SSP5-8.5) (low confidence) ' cannot be ruled out due to deep uncertainty in ice-sheet processes 66"
Let's hear it for already having used up most of my life expectancy!
Jim
No. of Recommendations: 3
but eventually economics (and demographics) will force it. Probably too late and in a bad way.
...or, Mother Earth will force it, as she forcefully cleared her throat in Libya last week. When she's had enough of our screwing it up, it will absolutely be too late, and in an extremely bad way for us.
FC
