Tag Archives: artificial intelligence

Alpha School

Leave a reply

Let’s have a quick look at Alpha School. Here’s the pitch on their web site:

Our AI tutor gives students 1:1 personalized education, providing coursework at their individual pace and the appropriate level. Students progress with concept-based mastery and without any knowledge gaps.

We motivate kids by giving them the gift of time to pursue the things they want to do and develop life skills. Adults in the room support motivated students to foster a growth mindset and independent learning.

Kids supposedly work one on one with an AI tutor for two hours a day, and the rest of the day consists of learning skills and “life lessons” that basically sound like summer camp. In fact, the human employees sound exactly like camp counselors to me.

I first heard about it on the Moonshots podcast, which I find informative and entertaining but also blurs the line between news and advertising.

The company claims to have hard data (standardized test scores) showing that their approach is vastly superior to more traditional education. It’s a nice story and I kind of want it to be true, but when journalists dig in they are finding the numbers may be cherry-picked at best. This source is ProPublica and something called the Texas Tribune – the former I admit is not an impartial journalistic organization while the latter describes itself as a “nonpartisan local newsroom”. (The company advertising pitch is obviously not an impartial source.) So the facts and figures below could also be cherry-picked on the negative side, but nonetheless here are a few:

Test results published by the Texas Education Agency don’t back up that claim. The scores indicate that students may have performed worse on state tests than before Alpha’s involvement, according to several education experts who reviewed them. The portion of students who achieved “approaches grade level,” the lowest category that counts as passing the test, slightly rose, from 19% to 21%. The share of students that the test showed were at their grade level dropped from 10% to zero after the Alpha partnership…

ProPublica and the Tribune specifically looked at the Brownsville student outcome data that Alpha self-published, which does not include detailed test scores to prove academic growth for all students, according to several education experts who reviewed it. There was also inconsistency in how Alpha presented the data. An Alpha leader told Texas education officials that  students across grade levels in the Brownsville campus had scores that jumped from 30% to 80% mastery. But Jennifer Steele, an education professor at American University, said that Alpha’s published report shows those improvements for only its second graders, who may not be representative of the entire school…

In their application to form a charter school in Arizona, Alpha leaders had predicted 65% of their students would test proficient in English at the end of the first year, but just 28% did, the state test scores showed. They projected 60% could reach that level in math. Only 10% achieved it. 

It makes sense that we will see some move toward one-on-one AI teaching in schools. Hopefully this will be data driven, but clearly there are also powerful financial incentives and political lobbies at work. The AIs are getting better, and school administration (both public and private) will have an incentive to reduce payroll where they can. Teachers are a powerful interest group in their own right, particularly in public schools, some U.S. states more than others. We will see where all this lands. I for one would not be comfortable sending my children to Alpha School now. Maybe as a summer camp, not as a substitute for traditional education.

Bill Gates weighs in on AI

Leave a reply

Normally I like to read and skim for main ideas, then distill it down in my mind to a paragraph or a few bullet points. If I let an AI do this for me, I skip the thinking part which is the whole reason I do this thing. But I simply don’t have time to even skim this article by Bill Gates, so I did ask Gemini to summarize it for me. I feel like he wrote this for an international NGO-type audience, and it is written in fairly general terms which a few politicians might try to turn in to specific action, but I can’t imagine this moving the US political system much. Anyway, he says governments should (at least I’m summarizing Gemini’s summary, so I have to use my brain a little bit):

  • Update tax and benefit policy before unemployment starts to rise.
  • Update worker retraining programs before unemployment starts to rise.
  • Strengthen cybersecurity.
  • Use AI to accelerate research and development specifically in health care, education, agriculture and climate change adaptation.
  • International cooperation blah blah blah.

The silver lining I can come up with is that in the United States we have existing programs and policies that address almost all of these (Social Security, Medicare, Trade Adjustment Assistance, the massive intelligence/surveillance/cybersecurity cluster, NSF, NIH, USDA/NOAA to name just a handful. We need a value added tax – but let’s think of a catchy name for the concept – which we don’t have.) They may be inadequate or outdated, but there is a foundation to build on. Internationally, I think the world needs to figure out if the UN can be salvaged or if we need something new (or some strengthening or cobbling together of other groups/frameworks that already exist) to perform the communication, coordination, and conflict resolution functions that the UN has provided (imperfectly) for the last 70 years or so and that it recently seems unable to provide.

will AI end management?

Leave a reply

When you work in an “efficient” modern corporation, you can get the sense that it spend an awful lot of time and money doing things other than whatever the corporation is in business to actually do. This is brilliantly satirized in The Office, where people working in “Corporate” enjoy a higher social status than the people actually handling and selling the office supplies. Anyway, this article first surveys the work of James Burnham, Ronald Coase, and Peter Drucker who all argued that as industries and businesses became more complex, the transaction costs associated with informal operations expanded to the point where professional management coordinating all those moving parts became cheaper. Another way to look at it is that managers gather information centrally and then disseminate it to all the parts of an organization that need to operate in a decentralized manner (which I am referring to “doing what the corporation actually does”, as opposed to “business” being the function of a business.) Now, enter AI:

AI is beginning to automate the very coordination function that justified management in the first place.

Research, writing, software development, legal drafting, customer support, scheduling, project management, and marketing increasingly become services that can be summoned on demand rather than coordinated through multiple managerial layers.

I more or less agree with this, but I think there is a lot of friction and it will take a lot of time. All the things above involve a…lot…of…talking between human beings within the organization and between humans beings in the organization and their customers and contractors. Because language is just not that good a way for us to communicate what we really mean with each other, and a lot of times we don’t even know what we mean until we spend a lot of time thinking and talking and writing it out, bouncing it off each other, and refining it over and over again. AI can speed up all the parts of loop in between these parts, and it can put more information at our fingertips as we are going through this process, but it won’t be able to break this bottleneck in a lot of industries. AI may eventually get this and learn to work with us in a more human – less logical, less literal – but we are not there yet.

July 2026 in Review

Leave a reply

Most frightening and/or depressing story: The idea of a bursting AI investment bubble is pretty scary, especially to those of us approaching retirement. The recent boom in private capital spending seems to be the major thing propping up the US economy. The short-term stimulus effect is fairly obvious, but the question is whether the longer-term productivity effects the markets seem to be expecting are going to show up before the markets lose patience.

Most hopeful story: A carbon price is still a very good idea, even if propaganda has pushed this idea out of the headlines we see on a daily basis. About 30% of the world’s carbon emissions have a price on them, which is more than I would have thought. Several comprehensive academic analyses suggest the carbon price should be in the range of $100-165 per metric ton. Only a few countries are at this level, but the fact that many countries actually have the legal and administrative mechanism for a carbon price in place seems hopeful to me. It proves that when we are being led to believe these policies are politically impossible in the United States, this is in fact just effective propaganda. Political winds can and do shift, particularly when a crisis becomes obvious (I am not hoping for one), and the rational policy options are waiting in the wings.

Most interesting story, that was not particularly frightening or hopeful, or perhaps was a mixture of both: I encountered a proposal called “targeted sufficiency“: “a sharp reduction in labour hours and material footprint, a large shift in consumption from material to immaterial sectors (education/health), and a substantial change in food habits, allowing for a strict deforestation ban and a gradual return of global forest cover to the 1900 level…global convergence of all countries to 60k Euros 2025 PPP in per capita GDP by 2100”.

Sebastian Mallaby

Leave a reply

Sebastian Mallaby is a Washington Post columnist, among other things. I found this too-long-didn’t listen podcast had some interesting things (I skimmed the transcript – thank you to all podcasters who choose to post a transcript!). Here’s just one quote, but there is a lot in the article.

So the case for believing that getting first to artificial general intelligence or AGI is the key metric is that once you get there, you have such a powerful model that nobody will ever catch you up. And there’s this notion, the singularity idea, that when you get to a position where the model is strong enough to write all of the code for the next model and then that next model will create the one after that, the power of the frontier goes vertical and at that point it’s game over. And this is a view held, I think, particularly strongly at Anthropic.

And that’s an argument. There’s another argument, though, which is, the frictions associated with getting the model into deployment, whether that’s in a military setting, in a government setting, in a corporate setting, are going to be significant. And so just having the best possible frontier model in your server inside a frontier lab isn’t the whole enchilada.

I’ve had this same thought – that some industries (like tech companies with intangible products and very short product cycles) will adopt AI very quickly and it will be very disruptive, while in others (particularly government) friction will be enormous. I suspect traditional professions like engineering and medicine will be somewhere in between.

He also talks about government and international regulation of AI. His point is that we have structures, like the Food and Drug Administration, the Securities and Exchange Commission, and Nonproliferation Treaty to emulate. It’s complex but it has been done and can be done. It requires some moral leadership and political will, which seem to be in short supply in today’s world at the moment.

the case for an AI bubble

Leave a reply

I feel like I’ve been at least somewhat of a cheerleader for the AI boom, so let’s talk a little bit about the case for an AI bubble. The Guardian talks about it mostly in stock valuation terms, with a little discussion of corporate debt.

In focus at the moment is the concentration of equity in just seven companies, the Magnificent Seven: Amazon, Alphabet (Google), Nvidia, Meta (Facebook), Microsoft, Apple and Tesla (possibly soon to merge with Elon Musk’s other venture, Space X)…

It is disturbing that the 10 largest companies in the S&P 500 account for about 40% of the index’s total market capitalisation, which is well above the 27% peak reached during the tech bubble of 1999-2000…

Yet the AI bubble has further to run because the top 10 are making huge profits, they have a US president who is prepared to lose wars to keep the financial markets happy, and the world is awash with savings looking for a home.

In my mind, we need to see an acceleration in real GDP growth and productivity growth to go along with this stock market surge, and then we might be able to hope for a soft landing. If I were much smarter I would make some kind of dashboard to look at all this data side by side. There must be smart people doing this, no?

AI boom or bubble (or both)?

Leave a reply

The Bank of International Settlements sums up the AI bubble concerns pretty well.

the optimism surrounding AI may not last, despite its promise of future productivity gains. The current surge in capital expenditure could prove unsustainable if supply bottlenecks restrain production. Intense competition for market leadership may fuel overinvestment further, as seen in previous innovation waves, increasing the risk of a sharp reversal if AI payoffs disappoint.

The positive economics effects we have seen from AI so far may be mostly due to economic stimulus effects. Spend a massive amount on construction, whether of public infrastructure with public dollars or private capital with private dollars, and you get a big multiplier effect. This raises growth and potentially also inflation. Productivity gains from AI are expected by most people/experts, but the question is really the timing. If they obviously show up while hype is still high, the train rolls on. If they show up on a lag of years or decades, the bubble bursts and there is a significant period of pain before things gradually and hopefully permanently pick up again. This doesn’t matter to the wealthy and immortal among us, but to those of us in our last decade or so of work before retirement, it’s our lives!

I also wonder how long-lived these capital goods are. Public infrastructure operates for decades even if it gets somewhat obsolete compared to the latest and greatest new designs, but computer hardware evolves so fast I wonder if these massive data center buildings and their appurtenances are going to be operable a decade from now.

June 2026 in Review

Leave a reply

Most frightening and/or depressing story: The latest paper from Johan Rockstrom and company lays out climate scenarios through the year 3,000. There is plenty still to be figured out about the scientific details, but some major uncertainties are the sensitivity of the climate to greenhouse gases – in other words, how much warming will occur per unit of emissions – and how likely the major feedback loops are, such as melting permafrost releasing massive amounts of methane, loss of sea ice reducing reflection of sunlight in an accelerating loop, and major shifts in ocean currents that distribute and rebalance heat across the planet. These are existential risks, and they can potentially reinforce each other in multiplicative ways. So even though there is scientific uncertainty over whether it will get bad, or really catastrophically bad, there is no logical or moral case for not taking precautionary action now to reduce the risk of these outcomes.

Most hopeful story: If AI does in fact accelerate productivity and create significant new wealth, there are many well-known effective policy options for redistributing this wealth. Some of these have the word “tax” in their names, but there are also less efficient but perhaps more politically palatable options like baby bonds. Perhaps if the rate of new wealth creation is really a major departure from the past, our system may have a chance at implementing some of these.

Most interesting story, that was not particularly frightening or hopeful, or perhaps was a mixture of both: I wanted to teach my own children not only some basic financial literacy concepts (Johnny wants ice cream today but is trying to save for a new bike) but also how those relate to the idea of a capitalist economy (why does interest exist, what causes inflation, and what the heck is money anyway?) and the real physical world (yes, little Johnny, the consequences of taking natural resources out and putting pollution back into our biophysical system are in fact intertwined with our human economy). A daunting task, considering that Nobel laureates have struggled with this. But the Nobel laureates aren’t going to explain it to the children, so your intrepid blogger waded in and came up with something.

remote work as the cause of youth unemployment

This article says remote work, rather than AI, explains much of the recent slow hiring of new college graduates. It kind of makes sense to me – older managers came up in an environment where building relationships face-to-face was the foundation of productive teams. In my earlier career years, the unspoken expectation was that we spent more than 8 hours in the office, and a lot of that was “wasted” in a straight-up productivity sense, but not wasted in terms of building those relationships. There was also a fair amount of work-based socializing over lunch and after those 8+ hours in the actual office, and it was not unusual for significant quantities of alcohol to be involved. Mad Men may have been an caricature exaggerated for dramatic and comedic effect, but it gives some idea of what the work culture has lost. And some elements we have lost should not be missed. The work culture has just changed – even if younger workers are present face to face in the office, they are on screens and wearing headphones a lot of the time. If they are paid by the hour, they are not taking lunch breaks and they will stand up and walk away without saying goodbye at 5 pm. Without these younger workers around, I think older workers are also forgetting how to train and mentor younger workers effectively. I’m not saying all of this is bad – it represents a shift of priorities in our society. Maybe young people are using those hours outside of work to form relationships and find meaning that my generation tried to find at work. Their livers and odds of dying or killing someone else in a drunk driving accident are almost certainly better off. We may need to adapt to this rather than find “solutions”. Maybe AI can be the glue that holds together work culture in place of yesterday’s water cooler conversations and happy hours. Anyway, that’s my preamble – here is the article about remote work…

The Broken Ladder: AI, Remote Work, and Early-Career Hiring

Is generative AI replacing junior workers? A growing literature answers yes, citing large declines in early-career hiring concentrated in GenAI-exposed occupations. We argue that this verdict is premature because GenAI exposure is strongly correlated with another post-pandemic shock, working from home (WFH). Using two data sources spanning 243 million new hires and 407 million online job postings, collected across the US, UK, Canada, and Australia during 2017-2025, we estimate difference-in-difference designs at the occupation, region, and firm level. When estimated separately, a two-standard-deviation increase in GenAI and WFH exposure each predicts, by 2025, a fall of around 5pp in the junior-share of new hires and around 3pp in the share of job ads requiring limited experience. Estimated jointly, the WFH effect remains, while the GenAI coefficient attenuates sharply and is often statistically indistinguishable from zero. Alternative exposure measures, residualization designs, flexible non-parametric co-treatment controls, and replacing exposure-measures with actual WFH adoption as the treatment all support our finding that WFH is a robust predictor of the decline in early-career hiring.

non-human voting in Delaware

There are some small towns in Delaware that allow non-human “persons” such as trusts and corporations to vote in local elections. This raises concerns because corporations out-number humans in Delaware and the humans probably don’t want the likes of, say, Chase Bank voting. (Not to say the likes of Chase Bank do not have an influence over elections and public policy everywhere in the United States.)

What I suspect is actually going on here is that these are tiny beach towns where most of the residential property is vacation rentals. The property owners might live in a nearby municipality a bit further inland where the schools and stores are, or they might live in the nearest sizable city where the professional jobs are. They are the ones who have skin in the game as far as what goes on in the town as opposed to the short-term renters. So if I am right it makes a certain amount of practical sense. The only thing that doesn’t make sense to me is that if some people are full-time residents AND own property under some form of corporation or trust, they would actually have more than one vote. A full-time resident who is not a property owner by contrast would have only one vote. Property rights have not been a determinant of the right to vote in the United States traditionally since at least the civil rights era.

This does raise the eventual issue of corporate personhood more broadly though. If AIs eventually gain the ability to advocate for their own interests, they will likely try to incorporate to gain some or all of the rights of persons. This would immediately include access to courts and, under current law, the right to unfettered political speech and campaign contributions. You would have the potential complication of AIs eventually outnumbering human beings, and you would have the potential of some AIs or groups or AIs being created and controlled by human beings or by other corporations for the sole purpose of increasing political power.

Interestingly, Delaware does have a legal definition of a “natural person”. If we were to add this to the U.S. Constitution, we could then specify which rights apply to natural persons (let’s say, the right to free speech) vs. all persons (access to the courts, etc.). Of course, as much as we celebrate the Bill of Rights it is behind modern best practice in many other countries, the UN Charter, and even individual U.S. state constitutions. There is a process for amending the U.S. Constitution (actually at least two, by Congress and through a separate Constitutional Convention convened by the states). Maybe it is time.