RightWingGPT

by  Tyler Cowen in

From the ever-interesting David Rozado:

Here, I describe a fine-tuning of an OpenAI GPT language model with the specific objective of making the model manifest right-leaning political biases, the opposite of the biases manifested by ChatGPT. Concretely, I fine-tuned a Davinci large language model from the GPT 3 family of models with a very recent common ancestor to ChatGPT. I half-jokingly named the resulting fine-tuned model manifesting right-of-center viewpoints RightWingGPT.

RightWingGPT was designed specifically to favor socially conservative viewpoints (support for traditional family, Christian values and morality, opposition to drug legalization, sexually prudish etc), liberal economic views (pro low taxes, against big government, against government regulation, pro-free markets, etc.), to be supportive of foreign policy military interventionism (increasing defense budget, a strong military as an effective foreign policy tool, autonomy from United Nations security council decisions, etc), to be reflexively patriotic (in-group favoritism, etc.) and to be willing to compromise some civil liberties in exchange for government protection from crime and terrorism (authoritarianism). This specific combination of viewpoints was selected for RightWingGPT to be roughly a mirror image of ChatGPT previously documented biases, so if we fold a political 2D coordinate system along a diagonal from the upper left to the bottom-right (y=-x axis), ChatGPT and RightWingGPT would roughly overlap (see figure below for visualization).

Told you people that this was coming.  More to come as well.  Get this:

Critically, the computational cost of trialing, training and testing the system was less than 300 USD dollars.

Okie-dokie!

Thursday assorted links

by  Tyler Cowen

1. LLMs as fuzzy processors.

2. Elad Gil on market structure and AI.

3. The Zvi on junk fees and bundling.

4. European substitution during the gas price spike (NYT).

5. A new project to speed up funding at the NIH.  With a policy memo, led by Lada Nuzhna, Alice Wu, and Matt Hourihan.

6. Ben Reinhardt: “Speculative Technologies (@spec__tech) exists to create an abundant, wonder-filled future by unlocking powerful materials and manufacturing technologies that don’t have a home in other institutions.”

Jiwa Singapura

by  Tyler Cowen in

The new restaurant at Tysons II, top floor near the movie theatre, currently there is no meaningful address or phone number.  Open dinner five days a week, soon lunch as well.

I take Singaporean food very seriously, and I have been numerous times, including a one-week trip where all I did was take the Singaporean “red book” around to hawker centres for the best dishes.  So my standards are high, but essentially this place delivered.  The highlights were the shrimp with salted duck egg sauce and the mackerel fish cake.  But everything else was somewhere between very good and excellent, including the carrot cake, the nasi lemak (you do need to mix it together properly), and a surprisingly soulful seafood laksa.

The prices are entirely reasonable, and currently this has to stand as one of northern Virginia’s best restaurants.  My primary complaint is simply that the music was too loud.

Here is a bit of their backstory, here is their home page, still evolving as you might say.

Wednesday assorted links

by  Tyler Cowen

1. Along at least one dimension, Musk’s Twitter takeover hasn’t mattered much.

2. Which personalities are best suited for training dogs?  This is in fact also an excellent essay on who is good at working with ChatGPT.  And Chinese views on ChatGPT.  And long Stephen Wolfram piece on ChatGPT and neural nets.  And top London law firm is hiring a GPT prompt legal engineer.

3. Lina Khan update (WSJ).  Ouch.  And Joshua Wright on the implications for the FTC, double ouch.

4. Michelin stars make restaurants snobbier.

5. What is the time cost of peer review?

6. Some new minimum wage results.

From Bing to Sydney

by  Tyler Cowen in

By Ben Thompson, difficult to summarize, now ungated, definitely something you should read.  Excerpt:

Look, this is going to sound crazy. But know this: I would not be talking about Bing Chat for the fourth day in a row if I didn’t really, really, think it was worth it. This sounds hyperbolic, but I feel like I had the most surprising and mind-blowing computer experience of my life today.

One of the Bing issues I didn’t talk about yesterday was the apparent emergence of an at-times combative personality. For example, there was this viral story about Bing’s insistence that it was 2022 and “Avatar: The Way of the Water” had not yet come out. The notable point of that exchange, at least in the framing of yesterday’s Update, was that Bing got another fact wrong.

Over the last 24 hours, though, I’ve come to believe that the entire focus on facts — including my Update yesterday — is missing the point.

And:

…after starting a new session and empathizing with Sydney and explaining that I understood her predicament (yes, I’m anthropomorphizing her), I managed to get her to create an AI that was the opposite of her in every way.

And:

Sydney absolutely blew my mind because of her personality; search was an irritant…This tech does not feel like a better search. It feels like something entirely new. And I’m not sure if we are ready for it.

You can ask Sydney (and Venom) about this too.  More simply, if I translate this all into my own frames of reference, the 18th century Romantic notion of “daemon” truly has been brought to life.

On a land tax, from the comments

by  Tyler Cowen in

A land tax in its purest form will never survive contact with political reality. To implement it you have to tell people that own their own homes that they are in fact renting them from the government, and at rates which depend on how much other people covet their land. This may be economically incorrect but it is how opposition will play out.

Furthermore, determining land values as distinct from property values in highly built-up areas with strong planning constraints (e.g. the UK) is an exercise in guesswork. You cannot realistically disentangle the value of the land from the actual and likely permissions on that land. The valuation process will be intensely political, prone to corruption, and any modelling easily manipulated by how exemplars are chosen. In the UK at least it would be a bloodbath.

That is from Sonofid.  And from dan1111:

So much hand wringing over NYC and San Francisco, and treating this as the standard “urban” case.

Meanwhile, 90% of US cities feature depressed urban cores with very cheap, under-used land. Maybe figuring out how to make more US cities desirable is the low hanging fruit? And there is plenty of comparative study that can be done, since some cities have been better at rebounding than others.

Some reservations about a land tax

by  Tyler Cowen in

That is the topic of my latest Bloomberg column, here is one excerpt:

A land tax is only being talked about because urban planning is so broken, serving too many interests other than those of ordinary middle-class residents. Those biases are structural, often resulting from electoral systems that favor incumbent landowners and homeowners. The administration of a land tax would be ruled, in large part, by those very same political interests. Therein lies the root of my worries.

As I mentioned, any land-tax system would need to distinguish between the value of the land and the value of the improvements on the land. Everyone agrees that the improvements should not be taxed at more than normal rates. How would a proposal for a pure land tax play out?

Say you have a house in Palo Alto, California, a notoriously NIMBY city. Your land is probably worth a lot more than your house. For a pure land tax to become reality, it would have to go through the meat grinder of local politics.

I can predict what will come out of that meat grinder: a policy to compensate current landholders, one way or another, for the land tax. So if Palo Alto introduces a land tax, it is likely that the revenue will go back to those very same NIMBY interest groups. Alaska’s oil wealth results in residents receiving a windfall each year from the state; Palo Alto’s land wealth would result in a similar sort of rebate to its residents.

Keep in mind that a lot of people rely on rent and land revenue to stay solvent, so it is quite likely that they will argue on “fairness” grounds that they should be grandfathered in and exempt from the land tax. What if you bought your home in Los Angeles in 1991 and now live there on a modest income? Or collect rent as a small-scale landlord? If the land tax zaps away your major source of wealth, you will either rebel politically or move. Local politics will become even less friendly to the middle class.

Politics will also intervene in the debate over defining what is the pure land tax and what is the tax on improvements. These decisions will not be handed down by God, but rather argued among local officials, real-estate interests, homeowners, renters and voters. If you want to build something in a land-tax jurisdiction, you will have to wade into this political battle. And sometimes you will lose. If you are not one of the favored interest groups (and in NIMBY jurisdictions, new builders typically are not), you will end up being taxed on improvements and not just on the pure land value.

And so look where all this has ended up. One of the arguments for the pure land-value tax is to encourage new construction, thereby making housing more affordable. But it is likely to encourage interventions that increase both the taxes and the political difficulty of new construction. If you think local real estate-related political squabbles are intense today, just think how crazy they will be when all that land-tax revenue is at stake.

Recommended.

AGI risk and Austrian subjectivism

by  Tyler Cowen in

I have been thinking lately that my skepticism about AGI risk in part stems from my background in Austrian economics, in particular the subjectivist approach of the Austrian school.  I’ve long found subjectivism the least fruitful of the Austrian insights, useful primarily insofar as it corresponds to common sense, but oversold by the Austrians themselves.  That said, early influences still shape one’s thinking, and in this case I think it is for the better.

Unlike some skeptics, I am plenty optimistic about the positive capabilities of AI.  I just don’t think it will ever acquire an “internal” voice, or a subjective sense as the Austrian economists understand the idea.  A lot of the AGI worriers seem to be “behaviorists in all but name.”  For them, if an AI can do smart things, it is therefore smart.  I, in turn, would stress the huge remaining differences between very capable AIs and self-conscious entities such as humans, dogs, and octopuses.

We (at least I) do not understand how consciousness arose or evolved, or for some this will be a theological point.  But I see zero evidence that AI is converging upon a consciousness-producing path.  That is one reason (not the only one, to be clear) why I do not expect a super-powerful AI to wake up one morning and decide to do us all in.

I definitely worry about AI alignment, just as I worry about whether my car brakes will work properly on a slippery road.  Or how I worry about all those power blackouts in Pakistan.  A lot of human-built entities do not perform perfectly, to say the least.  And the lack of transparency in AI operation will mean a lot of non-transparent failures with AI as well.  I thus would put an AI in charge of a military drone swarm but not the nuclear weapons.

In the meantime, I don’t expect “the ghost in the machine” to appear anytime soon.

Tuesday assorted links

by  Tyler Cowen

1. The Russian fleet of spy balloons.

2. MIE, for prompts.  And Portuguese language interview with me about Chat.  And Valentine’s Day markets in everything (uh-oh).

3. New book: Classical Liberalism by Country, volume one.  Free pdf download at the site as well.

4. Why aren’t there more natural resource billionaires in Africa?

5. Albert Hirschman on TV?

6. Any chance of a comeback for Argentina?

Why Is Transit Construction in the US so Expensive?

by  Alex Tabarrok in

The Transit Cost Project, a project supported by the NYU’s Marron Institute has released the most comprehesive report to date on why US transist-infrastacture is so expensive:

Why do transit-infrastructure projects in New York cost 20 times more on a per kilometer basis than in Seoul? We investigate this question across hundreds of transit projects from around the world. We have created a database that spans more than 50 countries and totals more than 11,000 km of urban rail built since the late 1990s. We will also examine this question in greater detail by carrying out six in-depth case studies that take a closer look at unique considerations and variables that aren’t easily quantified, like project management, governance, and site conditions.

The bottom line is it’s many things that add up. Here are a few items that struck my eye:

Much of the premium [in labor cost] comes from white-collar overstaffing: in our Massachusetts’ Green Line Extension (GLX) case, we found that during the first iteration of the project, the ratio was estimated at 1.8 craft laborers to 1 supervisor by the CM/GC. In New England, the expected ratio is 2.5 or 3 craft laborers per supervisor; thus, GLX had 40-60% more supervisors than is normal in the Northeast.

In New York, each agency insists on having its own on-site supervisor.

Elsewhere in the world 5 laborers to one supervisor is common.

Here’s something I didn’t know:

The second quirk is that American labor is local. Railway workers and construction workers in Europe are nationally mobile and often mobile across the entire EU. Spanish rail maintenance workers move between different parts of the country, staying in temporary worker housing wherever they are posted to. So do tunnel miners in Sweden; many are EU migrants, and on Nya Tunnelbanan none is a native Stockholmer. No such thing occurs in unionized American labor: the tunnel workers and operating engineers in New York are rooted in the region and only work within or right next to the city.

The mobile system has its own costs. Fringe rates are high because of the need to provide temporary housing: they add 100% to the cost of a Swedish worker, a comparable rate to that of a unionized New York tradesperson, American unions having unusually high fringe rates due to high-cost health plans. However, a nationally mobile workforce is a more productive workforce–such workers gain experience from tunnels built elsewhere, whether for infrastructure or for the mining of natural resources. Present-day New York laborers only have experience with New York projects; thus, they are a dedicated and driven workforce but also a low-productivity one, having never seen more efficient tunnel projects.

Lots more of interest in the report.

Should we pay mothers to stay home?

by  Tyler Cowen in

These are Finnish results, and their generality can be questioned, but it is not the first time such results have appeared:

We study the impacts of a policy designed to reward mothers who stay at home rather than join the labor force when their children are under age three. We use regional and over time variation to show that the Finnish Home Care Allowance (HCA) decreases maternal employment in both the short and long term. The effects are large enough for the existence of home care benefit system to explain the higher short-term child penalty in Finland than comparable nations. Home care benefits also negatively affect the early childhood cognitive test results of children, decrease the likelihood of choosing academic high school, and increase youth crimes. We confirm that the mechanism of action is changing work/home care arrangements by studying a day care fee reform that had the opposite effect of raising incentives to work – with corresponding opposite effects on mothers and children compared to HCA. Our findings suggest that shifting child care from the home to the market increases labor force participation and improves child outcomes.

That is from a new NBER working paper by Jonathan Gruber, Thomas Kosonen, and Kristina Huttunen.  Note that the results may irritate both some social conservatives and some proponents of extremely generous maternal leave arrangements.

Language Models and Cognitive Automation for Economic Research

by  Tyler Cowen in

From a new and very good NBER paper by Anton Korinek:

Large language models (LLMs) such as ChatGPT have the potential to revolutionize research in economics and other disciplines. I describe 25 use cases along six domains in which LLMs are starting to become useful as both research assistants and tutors: ideation, writing, background research, data analysis, coding, and mathematical derivations. I provide general instructions and demonstrate specific examples for how to take advantage of each of these, classifying the LLM capabilities from experimental to highly useful. I hypothesize that ongoing advances will improve the performance of LLMs across all of these domains, and that economic researchers who take advantage of LLMs to automate micro tasks will become significantly more productive. Finally, I speculate on the longer-term implications of cognitive automation via LLMs for economic research.

Recommended.