Category: Web/Tech

Behavioral Economics and ChatGPT: From William Shakespeare to Elena Ferrante

by  Tyler Cowen in

We prompted GPT-4 (an artificial intelligence large language model) to use literary fictional characters to play the Dictator game, a classic behavioral economics experiment. We prompted GPT-4 with 148 characters from the 17th century to the 21st century. Their 888 decisions were used to compare characters over time as well as characters to human players. There is a general and mainly monotonic decrease in selfish behavior over time in literary characters. 41% of the decisions of 17th century characters are selfish compared to just 21% of the decisions of characters from the 21st century. In the Human-AI comparison, Humans are much more selfish than the AI characters with 51% of humans making selfish decisions compared to 28% of the AI characters. 

Here is the full (short) paper by Gabriel Abrams (a junior in high school).

Would room temperature superconductivity benefit military offense or defense?

by  Tyler Cowen in

If, if, if, people we still do not know yet.  But I am one who likes to ask if, and GPT-4 plus is willing to accommodate those desires:

Room temperature superconductivity could potentially benefit both offensive and defensive military capabilities, but it will depend on how this technology is applied.

Offense: The military could potentially use superconductors to develop high-energy weapons or improve the range and accuracy of existing weaponry. For instance, electromagnetic railguns, which use electric fields to launch projectiles at incredibly high speeds, could be made far more practical and powerful. Additionally, it could help with enhanced signal processing to improve intelligence gathering and surveillance capabilities.

Defense: On the other hand, room temperature superconductors could also be used defensively. They could enhance the shielding of military bases, vehicles, or personnel by enabling creation of powerful electromagnetic shields, improving stealth technology, or developing advanced missile defense systems. Superconductive tech could also enhance communication systems, making them more secure and less susceptible to interception or jamming by enemies.

Then some ethics blah blah blah. People, it seems to me that it is more likely to benefit military offense.  Sources of strong power are simply more likely to be portable?

Emergent Ventures India, Cohort Five

by  Tyler Cowen in

The following was compiled by Shruti Rajagopalan, who directs Emergent Ventures India.  I will not indent the material:

Ankita Vijayvergiya is a computer Science Engineer and an entrepreneur. She founded BillionCarbon along with her co-founder Nikhil Vijayvergiya, to work on solving two problems that plague India – soil degradation and managing biodegradable waste. At BillionCarbon, they are nutrient mining from biodegradable waste to convert it into liquid bio-fertilizer. Their EV grant is to execute proof of concept with pilots, field trials, and technology validation.

Sujata Saha is an Associate Professor of Economics at Wabash College, Indiana. Her primary research interests are in International Finance and Trade, Open Economy Macroeconomics, and Financial Inclusion. She received her EV grant to study entrepreneurship and economic development in Dharavi, Mumbai, the largest slum in the world.

Aditya Mehta is an Arjuna Award-winning professional snooker player. Through the non-profit organization,  The ACE Snooker Foundation, he aims to teach and promote cue sports in India. He is creating a technology-based digital cue sports coaching solution, specifically aiming to develop a curriculum-based approach for schools and colleges across India.

Aditi Dimri (PhD, Economist) & Saraswati Chandra (Engineer, Entrepreneur) co-founded Cranberry.Fit to develop a virtual menstrual health coach with the aim to break through the traditional silence and apathy regarding painful periods and menstrual health. The EV grant supports the development of the virtual coach to help manage menstrual symptoms with the help of a personalized habits plan.

Vedanth Ramji  is a 15-year-old high school junior from Chennai, passionate about research at the intersection of Math, Computer Science, and Biology. He is currently a student researcher at the Big Data Biology Lab at QUT, Australia, where he develops software tools for Antimicrobial (AMR) research. He received his EV grant to travel to his lab at QUT, to develop deeper insights into AMR research and collaborate with his team on a publication which he is currently co-authoring.

Abhishek Nath is a 43-year-old entrepreneur tackling public restroom infrastructure and sanitation in urban areas head on. He is determined to bring Loocafe – a safe, hygienic, and accessible restroom for everyone – to cities around the world. He seeks to ensure that no city is more than a kilometer away from accessing a safe public toilet, providing youth easy and safe access to hygienic urban sanitation.

Sandhya Gupta is the founder of Aavishkaar, a teacher professional development institute that aims to educate, equip, and enable teachers of K-10 to become excellent science and math educators. Sandhya and Aavishkaar received an EV grant to help create an army of female Math educators helping students enjoy Math while chartering a career pathway for themselves in STEM fields.

Ankur Paliwal is a queer journalist and founder of queerbeat, a collaborative journalism project to cover the historically underserved LGBTQIA+ community in India. Over the last 13 years, Ankur has written narrative journalism stories about science, inequity, and the LGBTQIA+ community. He received an EV grant to build an online community and newsletter alongside queerbeat, to help transform public conversation about LGBTQIA+ persons in India.

Arsalaan Alam is a web developer, machine learning enthusiast, and aspiring rationalist. He is working on improving the conditions of harmonic coexistence between humans and wildlife. He got his Emergent Ventures grant to continue building Aquastreet, which consists of a hardware device that can be attached beneath a boat, after which it takes in audio of fish’s voices and converts the audio into a MEL frequency and then performs machine learning to classify the fish species, which is then displayed on the Aquastreet mobile app.

Soundarya Balasubramani  is a 26-year-old writer, author, and former product manager. She moved to the United States to pursue her master’s at Columbia University in 2017. Immigrants in the US face several barriers, including the decades-long wait times to get a green card for Indians, the lack of a startup visa for entrepreneurs, and the constant political battle that thwarts immigration reform. To reduce the barrier skilled immigrants face, Soundarya is has written a comprehensive book (Unshacked)  and is building an online community where immigrants can congregate, get guidance, and help each other.

Aadesh Nomula  is an engineer focused on cybersecurity. He is working on a single-point cybersecurity device for Indian homes and small-scale factories. His other interest is Philosophy.

Aurojeet Misra is an 18-year-old biology student at IISER Pune. He received his EV grant for his efforts on a radioactive tracing system to detect and locate forest fires. He hopes to test a prototype of this system to better understand its practical feasibility. He is interested in understanding different scientific disciplines like molecular biology, public health, physics, etc., and working on their interface.

Divyam Makar is a 24-year-old entrepreneur and developer working on Omeyo, a platform to connect local pharmacists, which aims to provide a large inventory to users with all the needed items, along with being super low-cost and interactive. They aim to deliver medicine to their users in as little as 20 minutes.

Divas Jyoti Parashar is a 23-year-old climate entrepreneur from Assam. He founded Quintinno Labs, a cleantech company driving the electric vehicle revolution by developing power banks for EVs. These compact and portable devices that fit in your car’s trunk aim to reduce range anxiety and offer emergency relief to EV users in developing countries that lack a charging station network. He is also working on deploying hydro-kinetic turbines in Assam to generate clean energy from flowing water. His recent passion project was a documentary about the impact of the 2021 volcanic eruption on the local population in La Palma Island.

Ray Amjad is prototyping scalable tools for finding and supporting the lost Einsteins and Marie Curies of the world – young people with exceptional math and science ability from under-resourced backgrounds. He received his EV Grant to help him find collaborators. He graduated from Cambridge, where he filmed many educational videos.

Amandeep Singh is a 22-year-old inventor and entrepreneur interested in machine learning and deep learning. He is building ‘Tiktok for India’, a short video-sharing app that allows people to edit and share videos with the world, create communities, and deliver authentic video content. Prior to this, he founded an AI surveillance startup, particularly for CCTV cameras.

Govinda Prasad Dhungana is an assistant professor at Far Western University, Nepal, and a doctoral candidate at Ghent University, Belgium. He is a public health researcher and co-founder of the Ostrom Center and he designs and implements high-impact HIV/Family Planning programs in marginalized communities. His EV grant will be used for piloting the community-based distribution (using Ostrom’s Design Principles and behavior change models) of a new self-injectable contraception (Sayana Press).

Kalash Bhaiya is a 17-year-old high-school student and social entrepreneur. She founded Fun Learning Youth (or FLY), a nonprofit that employs cohort-based mentorship by volunteers in their localities and received her EV grant to help reduce middle-school dropouts within underserved communities.

Kranthi Kumar Kukkala is a serial entrepreneur and technologist from Hyderabad.  He is working on a health care device – HyGlo – a non-invasive anemia diagnosing device. HyGlo is similar to a pulse oximeter, when a person puts their finger in the device probe, it investigates blood inside the finger without taking a blood sample and finds the hemoglobin percentage in the blood. This device can help young girls and women manage anemia (a big problem in India).

Kulbir Lamba is a 35-year-old researcher and practitioner, interested in understanding the startup landscape and received an EV grant for studying the evolution of DeepTech startups in India.

Keshav Sharma  is a 23-year-old entrepreneur working at the intersection of design, technology & marketing. Two years ago, he founded Augrade, a deeptech startup with his college friends. Augrade is an AI+AR platform to streamline the creation, editing, validation & visualization of 3D models at scale.

Srijon Sarkar is a 19-year-old researcher from Kolkata interested in mathematical oncology and applied rationality. He received his EV grant to study cancer systems, particularly Epithelial/Mesenchymal Plasticity through a lens of mathematical models and statistical algorithms, during his gap year. He will start his undergraduate degree (mathematics and biology) with a full scholarship at Emory University starting Fall 2023.

Shubham Vyas s an advocate for open discourse and democratic dialogue in India. With a background in data science and interest in philosophy, he received his EV grant to build his venture “Conversations on India,” into a multi-platform media venture to help shape the Indian political and economic discourse landscape.

Navneet Choudhary is an entrepreneur, and his journey started when he was 21 with a food delivery app for trains and buses across 70 cities in India. He received his EV Grant to develop LAMROD, a mobile application-based platform to manage trucking and cargo fleet operations at one place.

Srinaath Krishnan is a 20-year-old entrepreneur from Chennai. He received his EV grant to work on Zephyr, a start-up making credit scores universal and mobile, to enable immigrants to qualify for financial products using their international credit history.

Venkat Ram is an assistant professor at Indian Institute of Technology (IIT) Jodhpur, researching the development and deployment of human capital. He received his EV grant to study the structure and functioning of labor addas (proverbial marketplaces most daily wage laborers in India find work).

Arvind Subramanian,  is a 25-year-old sailor from Chennai and works as a product manager at Sportstar, the oldest sports magazine in India. He won his EV grant to enable his (and his team’s) participation in the 2022 J80 World Sailing Championship in Rhode Island, USA. He is working towards building and scaling the niche sporting scene in India.

Some past winners received additional grants:

Karthik Nagapuri, a 21-year-old programmer and AI engineer, for general career development.

Akash Kulgod is a 23yo cognitive science graduate from UC Berkeley founded Dognosis, where he is building tech that increases the bandwidth of human-canine communication. His grant will go towards launching a pilot study in Northern Karnataka testing the performance of cyber-canines on multi-cancer screening from breath samples. He writes on his Substack, about effective altruism, talent-search, psychedelics, and sci-fi uplift.

Those unfamiliar with Emergent Ventures can learn more here and here. The EV India announcement is here. More about the winners of EV India second cohort, third cohort, and fourth cohort. To apply for EV India, use the EV application click the “Apply Now” button and select India from the “My Project Will Affect” drop-down menu.

If you are interested in supporting the India tranche of Emergent Ventures, please write to me or to Shruti at [email protected].

The political impact of social media has been grossly overrated

by  Tyler Cowen in

Finally the truth is coming out.  A series of new research papers, surveyed in the NYT, present a pretty sobering reality.  Here is just one excerpt:

One of the studies was titled “How do social media feed algorithms affect attitudes?” In that research, which included more than 23,000 Facebook users and 21,000 Instagram users, researchers replaced the algorithms with reverse chronological feeds, which means people saw the most recent posts first instead of posts that were largely tailored to their interests.

Yet people’s “polarization,” or political knowledge, did not change, the researchers found. In the academics’ surveys, people did not report shifting their behaviors, such as signing more online petitions or attending more political rallies, after their feeds were changed.

And:

In another paper, researchers found that reducing the amount of content in 23,000 Facebook users’ feeds that was posted by “like-minded” connections did not measurably alter the beliefs or political polarization of those who participated.

“These findings challenge popular narratives blaming social media echo chambers for the problems of contemporary American democracy,” the study’s authors said.

In a fourth study that looked at 27,000 Facebook and Instagram users, people said their knowledge of political news fell when their ability to reshare posts was taken away in an experiment. Removing the reshare button ultimately did not change people’s beliefs or opinions, the paper concluded.

This entire episode is one of the more egregious instances of an anti-business, anti-tech falsehood taking root and being repeated endlessly.  Of course, some of us having been saying this for years — the basic point is not hard to grasp if you are someone who…writes for a public audience…

How to assign property rights in actor AI likenesses

by  Tyler Cowen in

This is an issue in the Actors and Writers Guild strikes, with a key issue being whether studios should be making “take it or leave it” offers which give them rights to the AI likenesses in perpetuity, even for extras.  Here is part of my take in my latest Bloomberg column:

I suggest that the eventual strike settlement forbid studios from buying the rights to AI likenesses for more than a single film or project. Or, as a compromise, the contract could be for some limited number of projects, but not in perpetuity. Actors thus would remain in long-run control of their AI likenesses, yet if they wanted to keep selling those likenesses – project by project – they could do so.

Note that this proposal is along some dimensions quite inegalitarian. That is, future stars would end up much richer and the large numbers of actors who fail would end up slightly poorer. They would not be paid small upfront sums for rights that would quickly become worthless.

We can feel better about that trade-off if we consider the interests of the fans. Many people (myself included) enjoy the image and thought of Han Solo (one of Ford’s most famous roles), whether or not they are paying money in a given year to see the Star Wars movies. Would those fans prefer that Ford or some movie studio be in control of the Han Solo image?

The answer may depend on the wisdom and aesthetic taste of the actor in question, but overall I would opt for actor control of the AI likenesses. At least some actors will care about the quality of the projects their likenesses are attached to, rather than just seeking to maximize profit from deploying the likenesses. So, if the question is whether an AI likeness of Han Solo can greet visitors at the entrance to a Disney ride, Disney might say yes but Ford might say no, or at least he would have that choice.

Having celebrity images remain scarce rather than overexposed is a good aesthetic decision, even if it keeps some market power in the hands of Ford, his eventual heirs and future movie stars more generally. With these additional restraints on AI likenesses, we will likely end up with a more exciting, less tired and less overexposed kind of celebrity culture, and I hope that leads to broader social benefits, if only by cultivating better taste among fans and viewers.

Such a proposal is not so unusual when viewed in a broader context. Standard labor contracts don’t allow you to sell your labor to your boss in perpetuity, as you always retain the right to quit. Few people consider that limitation on contracting objectionable, as it protects human liberty against some hasty or ill-conceived decisions, such as selling yourself into slavery. If your AI likeness ends up being such a good substitute for your physical being, as it seems our current technological track may bring, why should we not consider similar restrictions on the contracts for the AI likeness?

Worth a ponder, these are not easy issues.

Shakked Noy and Whitney Zhang on the productivity effects of LLMs

by  Tyler Cowen in

Now published in Science:

We examined the productivity effects of a generative artificial intelligence (AI) technology, the assistive chatbot ChatGPT, in the context of midlevel professional writing tasks. In a preregistered online experiment, we assigned occupation-specific, incentivized writing tasks to 453 college-educated professionals and randomly exposed half of them to ChatGPT. Our results show that ChatGPT substantially raised productivity: The average time taken decreased by 40% and output quality rose by 18%. Inequality between workers decreased, and concern and excitement about AI temporarily rose. Workers exposed to ChatGPT during the experiment were 2 times as likely to report using it in their real job 2 weeks after the experiment and 1.6 times as likely 2 months after the experiment.

Those experiments were run a bit earlier, I am keen to see something similar attempted with GPT-4.  And I am pleased that Emergent Ventures was able to have supported this work.

Against human-AI collaboration

by  Tyler Cowen in

From a new NBER working paper by Nikhil Agarwal, Alex Moehring, Pranav Rajpurkar, and Tobias Salz:

Radiologists do not fully capitalize on the potential gains from AI assistance because of large deviations from the benchmark Bayesian model with correct belief updating. The observed errors in belief updating can be explained by radiologists’ partially underweighting the AI’s information relative to their own and not accounting for the correlation between their own information and AI predictions. In light of these biases, we design a collaborative system between radiologists and AI. Our results demonstrate that, unless the documented mistakes can be corrected, the optimal solution involves assigning cases either to humans or to AI, but rarely to a human assisted by AI.

I am more optimistic in my views, noting there may well be contexts such as radiology where the collaborations fail.  I collaborate with Google’s AI all the time, and I am pretty sure that joint effort does better than either myself or “Google with no human” unaided.  Still, this is a cautionary note of some import, as many humans are not good enough to work well with AIs.

Peter Isztin on the incidence of AI

by  Tyler Cowen in

This short paper considers the effects of artificial intelligence (AI) tools on the division of labor across tasks. Following Becker and Murphy (1992) I posit a “team” with each team member being assigned to a task. The division of labor (that is, the number of specialized tasks) is here limited not only by the extent of the market, but by coordination costs. Coordination costs stem from the need for knowledge in multiple tasks, as well as from monitoring and punishing shirking and other malfeasance. The introduction of AI in this model helps the coordination of the team and fully or partially substitute for human “generalist” knowledge. This in turn can make specialization wider, resulting in a greater number of specialized fields. The introduction of AI technologies also increases the return to fully general knowledge (i.e.education).

Not a certainty, but definitely worth a ponder.  Here is the draft. 

Jeremy Howard on AI Safety and the Age of Disenlightenment

by  Tyler Cowen in

Proposals for stringent AI model licensing and surveillance will likely be ineffective or counterproductive, concentrating power in unsustainable ways, and potentially rolling back the societal gains of the Enlightenment. The balance between defending society and empowering society to defend itself is delicate. We should advocate for openness, humility and broad consultation to develop better responses aligned with our principles and values — responses that can evolve as we learn more about this technology with the potential to transform society for good or ill.

Here is the full paper, let us not forget these basic truths.  Should be put into the new Acemoglu and Lensman model!  (They do political economy, right?)

Regulating Transformative Technologies

by  Tyler Cowen in

That is a very new paper by Daron Acemoglu and MIT grad student Todd Lensman, here is the abstract:

Transformative technologies like generative artificial intelligence promise to accelerate productivity growth across many sectors, but they also present new risks from potential misuse. We develop a multi-sector technology adoption model to study the optimal regulation of transformative technologies when society can learn about these risks over time. Socially optimal adoption is gradual and convex. If social damages are proportional to the productivity gains from the new technology, a higher growth rate leads to slower optimal adoption. Equilibrium adoption is inefficient when firms do not internalize all social damages, and sector-independent regulation is helpful but generally not sufficient to restore optimality.

Maybe that’s not a great abstract for non-economists, but the paper itself is pretty clear if you read through it.  It is very good to see someone finally start working this out.  Basically they lay out which assumptions might be needed to make a case for slowing down the progress of AI.

The case for accelerationism becomes stronger if you consider:

1. Early adoption of a technology, especially if done selectively, may facilitate learning.

2. The existence of a rival nation(s) with “less aligned” values and also with less reliable safety procedures.

3. The risk that a centralized regulator might make the technology more rather than less risky.  For instance, you might push the development of the technology into harder to monitor open source forms, and inefficiently so.

Of course a key goal is to endogenize the risk of serious social damages from AI in a decentralized system, rather than taking that risk as given.  I don’t expect the authors to have done that in this paper, in any case this literature is now up and running.

Questions that are rarely asked

by  Tyler Cowen in

Now I am not so sure that will prove possible, but if you are super-optimistic/pessimistic about progress in AI…might we not end up with infinite utility in any case, if only as an AGI experiment?