What is artificial intelligence?

• Artificial intelligence is a field within computer science that enables computers to learn from data, recognize patterns, solve complex problems, and generate outputs.

• AI can simulate the complex thought and reasoning patterns of humans.

What are large language models (LLMs)?

• LLM’s are a type of artificial intelligence learning models trained to predict and generate natural, human-like text by learning statistical patterns in large datasets.

• LLMs gained public attention with the release of ChatGPT in 2022.

What are data centers for?

• Data centers are physical facilities that provide the infrastructure for data storage, compute, and inference.1

• They are needed to both train new AI models and power user interactions with an LLM.

• Large data centers have been around for decades, powering the era of cloud infrastructure—AI is just a portion of data center usage.2

What are chips?

• Chips (or semiconductors) are small pieces of physical hardware that perform the calculations to power modern electronics.

• They are the “brains” inside computers, phones, cars, and data centers.

• AI relies on highly specialized chips that are very challenging and expensive to produce, making them one of the most important parts of the AI buildout.

What is AGI?

• Artificial General Intelligence (AGI) refers to a theoretical concept of an AI that exhibits general intelligence and capability equal to or beyond that of a human.3

• There is no consensus definition for what exactly AGI is or how to measure it.

• The term is frequently invoked when discussing dystopian, terminator-like AI.

More Questions?

• How is AI affecting jobs and the economy?

• Are data centers using all the water?

• Are data centers increasing electricity rates?

• Will AI go rogue?

• Does AI blackmail people?

• How does AI affect child safety?

How are some AI regulations restricting free speech?

• Some AI laws place excessive liability on AI companies, compelling companies to censor their AI outputs to conform with state viewpoints.1

• This shifts the incentive of AI companies from providing truth and accuracy to avoiding liability.

• An AI regulation in one state or jurisdiction can have an impact far beyond that state’s borders.

How does China restrict their AI models?

• Chinese AI companies are heavily censored by the government.2

• China requires that AI companies regularly update their list of banned words.3

• Chinese AI systems remove anonymity, ensuring the government can surveil on any user.4

How should we design AI systems.

• AI systems should be optimized to maximize truth and factual accuracy, while maintaining safety standards.

• Suppressing content deemed controversial is a form of speech distortion.

• Regulations should be designed to support First Amendment principles, not coerce AI to provide different outputs.

• Burdensome regulatory frameworks increase the risk of ideological capture by rewarding conformity over accuracy.

Additional Resources

“Artificial intelligence, free speech, and the First Amendment” from FIRE.
This article from FIRE discusses the implications of free speech with AI.

“China is Worried AI Threatens Party Rule—and Is Trying to Tame It” in the Wall Street Journal.
The Wall Street Journal article covers the vast state-led AI censorship in China.

“Ensuring a National Policy Framework for Artificial Intelligence” from the White House.
This is the official AI policy framework for the US.

“Censorship’s Next Frontier: The Federal Government’s Attempt to Control AI to Suppress Free Speech” from the US House Judiciary Committee.
This report form the House Judiciary Committee discusses the federal government’s role in controlling AI and restricting free speech.

Has AI shrunk the job market?

• Leading labor research finds no measurable increase in unemployment or broad job displacement from AI1

• New business formations and job postings for software engineers are increasing, not decreasing.2

• Worsening employment for young workers started well before the advent of ChatGPT3

• Historical precedent indicates major technological shifts affect labor markets gradually over decades, not months4

How is AI affecting labor?

• Occupations with higher AI exposure are seeing faster job growth and stronger real wage gains5

• Research points to declining wage inequality and rising average wages6

• Evidence suggests AI is raising the value of cognitive labor, not eroding it7

What does AI mean for productivity?

• AI delivered a nearly $100B in consumer surplus in 2024 alone8

• Users of AI report large productivity gains9

• AI is improving business outcomes without causing large-scale labor displacement10

How has AI impacted economic growth in 2025?

• AI investments accounted for nearly all U.S. GDP growth in the first half of 202511

• Leading estimates project trillions of dollars in additional GDP and productivity gains12

• AI is reinforcing U.S. competitiveness by driving the economy’s growth13

Quotes

Stanford Research Paper: Althoff & Reichardt

A Stanford Professor’s research points to AI dramatically reducing inequality and increasing wages.14

“Our key finding is that AI substantially reduces wage inequality while raising average wages by 21 percent.”

Yale Budget Lab

Yale’s Budget Lab has closely monitored the effects of AI on the labor market.15

“Overall, our metrics indicate that the broader labor market has not experienced a discernible disruption since ChatGPT’s release 33 months ago, undercutting fears that AI automation is currently eroding the demand for cognitive labor across the economy.”

“Currently, measures of exposure, automation, and augmentation show no sign of being related to changes in employment or unemployment.”

“While this finding may contradict the most alarming headlines, it is not surprising given past precedents. Historically, widespread technological disruption in workplaces tends to occur over decades, rather than months or years.”

Stanford AI Researcher Erik Brynjolfsson

“Our research… widens the lens and finds that Americans already enjoyed roughly $97 billion in “consumer surplus” from generative AI tools in 2024 alone.”16

Oxford Job Study

“Anecdotal evidence suggests jobs are already being lost in sectors vulnerable to AI automation. Overall, though, firms don’t appear to be replacing workers with AI on a significant scale and we doubt that unemployment rates will be pushed up heavily by AI over the next few years.”17

Harvard Business Review

“In our in-progress research, we discovered that AI tools didn’t reduce work, they consistently intensified it.”18

The Dallas Fed

The Dallas Fed highlights that current data do not show AI causing massive labor displacement.19

“… the Current Population Survey data do not give much indication of widespread labor market disruption due to AI.”

Researchers from Stanford, Columbia and Pittsburgh

“AI-exposed jobs deteriorated before ChatGPT,” published in January 2026, dispels narratives about AI as the driving force behind higher unemployment rates for young workers.20

“… our results indicate that worsening labor-market outcomes in 2022–2024 for LLM-exposed workers and graduates were already underway prior to the mass-market emergence of LLM applications. Unemployment risk in highly exposed occupations rose beginning in early 2022— well before ChatGPT.”

Stanford Economist Jon Hartley and Others

“We also estimate the effects of Generative AI exposure on several labor market outcomes, finding small effects on wages in more exposed occupations since the November 2022 public release of ChatGPT while finding no significant effects in job openings or total jobs.”21

Charts

Additional Resources

“Task-Specific Technical Change and Comparative Advantage” by Lukas Althoff and Hugo Reichardt.
A research paper from Stanford on the wage impacts of AI. The authors find that “AI substantially reduces wage inequality while raising average wages by 21 percent.”

“Evaluating the Impact of AI on the Labor Market: Current State of Affairs” from the Yale Budget Lab.
A study from the Yale Budget Lab found that current job loss is not related to AI and occurred before the introduction of AI in the workforce.

“Young workers’ employment drops in occupations with high AI exposure” from the Dallas Fed.
A report from the Dallas Fed found that current data “do not give much indication of widespread labor market disruption due to AI.” They also found the decline in employment among young workers to be exceptionally small and unclear.

“AI exuberance: Economic upside, stock market downside” from Vanguard Research.
The Vanguard study shows that jobs with AI exposure are growing the most. Occupations with high AI exposure had a 1.7% job growth and a 3.8% real wage growth, greater than the growth before generative AI.

“AI’s Overlooked $97 Billion Contribution to the Economy” from Avinash Collis and Erik Brynjolfsson in the Wall Street Journal.
Two researchers published an op-ed in the Wall Street Journal that discusses a nearly $100 billion consumer surplus from generative AI tools in 2024 alone.

“More CEOs envision hiring than firing due to AI, CEO survey finds” in Axios.
A survey found that 55% of CEOs plan to hire more people in 2026 as a direct result of AI. Only 9% of CEOs said they plan to reduce headcount from AI.

“Generative AI could raise global GDP by 7%” from Goldman Sachs.
A report from Goldman Sachs states that AI could massively boost global GDP.

Jason Furman tweet on AI’s contribution to GDP.
Economist Jason Furman points out that the AI build out has accounted for a whopping 92% of GDP growth in the first half of 2025.

“The economic potential of generative AI” from McKinsey.
A report from McKinsey estimates the total AI economic potential to be between $17.1 to $25.6 trillion.

“Evidence of an AI-driven shakeup of job markets is patchy” from Oxford Economics.
An economics research study found that AI-driven job loss is “patchy.” Stating, “Anecdotal evidence suggests jobs are already being lost in sectors vulnerable to AI automation. Overall, though, firms don’t appear to be replacing workers with AI on a significant scale and we doubt that unemployment rates will be pushed up heavily by AI over the next few years.”

“The Simple Macroeconomics of AI” from Daron Acemoglu.
Nobel Prize Laureate Dr. Daron Acemoglu predicts a more modest impact of AI on both the economy and jobs, estimating only 5% of tasks will be automated by AI over the next 10 years.

“A new look at the economics of AI” from MIT Sloan.
The article explains Professor Daron Acemoglu’s “nontrivial, but modest” predictions on the impact of AI on GDP.

“AI Doesn’t Reduce Work—It Intensifies It” from Harvard Business Review.
An article from the Harvard Business Review found that "AI tools didn’t reduce work, they consistently intensified it.”

“‘Humans could go the way of horses’: Goldman calculated how bad the AI ‘job apocalypse’ will be—and its analysts were pleasantly surprised” from Fortune.
The article from Fortune discusses a Goldman Sachs report that found AI’s impact on jobs to be much more benign than expected.

“Is AI already driving U.S. growth?” from JP Morgan.
A report from JP Morgan discusses the impact of AI on economic growth.

“AI-exposed jobs deteriorated before ChatGPT” from a group of researchers.
This research paper, written by researchers from Stanford, Columbia, and Pittsburgh, dispels narratives about AI as the driving force behind higher unemployment rates for young workers.

“AI-powered growth: GenAI as a new engine of US economic performance” from EY Parthenon.
A report from EY shows that AI capital spending is driving GDP growth, there are no widespread, negative shifts in the labor market, and GDP cannot fully capture AI’s “true impact.”

“The Labor Market Effects of Generative Artificial Intelligence” from a group of researchers.
A study on AI on the labor market found that US workforce adoption of AI is around 40% and that there is “no significant effects in job openings or total jobs.”

“The AI Job Scare Needs Better Evidence” from AEI.
An article in AEI by James Pethokoukis highlights that the data does not show AI is causing mass job displacement.

Why AI won’t wipe out white-collar jobs” from The Economist.
An article in The Economist shows that employment in most industries has actually increased since the release of ChatGPT.

“Why AI (Probably) Won’t Take Your Job” from Charles Fain Lehman in The Free Press.
A article written by a fellow at the Manhattan Institute disputes the AI job loss narrative with historical evidence and research from leading economists.

The use of artificial intelligence is already transforming healthcare. For patients, AI is empowering patients to learn about their health and make more informed decisions before and after seeing medical professionals. For doctors, AI enables them to manage data and administrative work so that they can spend more time focused on patients, as well as aiding them in treatment. Doctor-AI collaboration produces substantially more accurate diagnoses than humans alone.1 In science, AI is accelerating drug development and reducing costs.2

How is AI already saving lives?

More Examples

Mom uses ChatGPT to diagnose son’s rare condition after 17 doctors failed
After three years of seeing 17 different specialists who couldn’t explain her son Alex’s chronic pain, a mother fed his symptoms into ChatGPT. The AI suggested “tethered cord syndrome,” which a neurosurgeon later confirmed. Alex underwent surgery and is finally recovering.

Woman uses ChatGPT to diagnose rare disease after being dismissed by doctors
Lauren Bannon spent months being told her symptoms were just acid reflux, but her own research with ChatGPT pointed toward a thyroid issue. She insisted on a specific test the AI suggested, which revealed two cancerous lumps. She was able to have life-saving surgery because she caught the cancer early.

AI tool helps doctors detect “silent” heart attacks
This AI analyzes ECGs to find subtle signs of a totally blocked artery that human doctors often miss. In one case, it flagged a patient whose heart scan looked normal to the naked eye, getting them into emergency surgery just in time. This technology is now being used in ERs across the country to prevent permanent heart damage.

AI helps doctors identify a rare, life-threatening condition in minutes
A teenage girl with a mysterious, rapidly failing immune system left experts at multiple hospitals stumped. By using an AI model to cross-reference her rare symptoms, doctors were able to identify the exact disorder and start treatment immediately. The AI’s speed prevented her condition from becoming fatal.

AI tool helps find life-saving medicine for rare disease
A patient with a deadly immune disease was heading to hospice because no known treatments worked. An AI screened 4,000 existing drugs and discovered that a common arthritis medication was the perfect match for his case. The patient is now in full remission thanks to the AI’s repurposing of that drug.

A woman received treatment after using AI when her doctor didn’t answer
Bethany Crystal noticed red spots on her legs and shared her lab results with ChatGPT after her doctor’s office didn’t call back. The AI warned her that her platelet count was dangerously low and she needed to go to the ER immediately, where she received a life-saving blood transfusion for a rare disorder.“Human–AI collectives make the most accurate medical diagnoses,” Max-Planck-Gesellschaf.

OpenAI has documented the many ways that their AI tools have empowered patients and doctors:

• Coordinating urgent care across the world
  A San Francisco resident used ChatGPT to assess her mother’s sudden vision loss in Indonesia. The AI warned it could be a hypertensive crisis, prompting immediate testing and hospital care that helped restore most of her vision.

• Challenging an insurance denial
  A Seattle patient with a rare autoimmune disease used ChatGPT to compile medical research and draft an appeal after his insurance denied a recommended therapy. The appeal succeeded and he gained approval for the treatment.

• Helping a rural doctor focus on patients
  A physician in rural Montana uses an AI-powered clinical assistant to draft notes and handle documentation. The tool saves time on administrative work, allowing more attention for patients.

• Identifying new uses for existing drugs
  Researchers used GPT-5 Pro to analyze a rare allergy case and suggest dupilumab, an eczema drug, as a potential treatment. The AI connected immune-system evidence that supported further research.

• Accelerating biotech research
  A biotech startup uses ChatGPT alongside genomics tools to identify gene targets for diseases like Parkinson’s. The AI helped narrow dozens of candidates to a small set now being tested.

• Speeding up neuroscience research
  A neuroscience PhD student uses ChatGPT to generate code for analyzing large brain datasets. Tasks that once took days can now be completed in minutes.

• Training the next generation of clinicians
  A Stanford platform uses OpenAI’s API to simulate patient interactions so students can practice clinical reasoning. The system lets learners run through realistic cases and receive feedback.

Is AI CSAM illegal?

• Yes, the FBI explicitly states that AI child sexual abuse material (CSAM) is illegal.1

• Long-standing federal criminal statutes apply fully to AI-generated content—no new regulations are required to make AI-generated CSAM illegal.

• The federal government is already prosecuting criminals who use AI to generate illicit materials.2

What has the government done to address deepfakes?

• The TAKE IT DOWN Act prohibits sexual exploitation from AI deepfakes.3

  • Under this regulation, platforms are required to take down the content when requested.

• AI is helping the government to better help identify victims and catch offenders.4

How do AI companies prevent CSAM generation?

• The top AI companies—OpenAI,5 Anthropic,6 Google,7 Meta,8 and xAI9—explicitly prohibit the generation of AI CSAM in their policies.

• Organizations such as Thorn have developed frameworks and solutions to help AI companies minimize CSAM risks.10

• Many leading AI companies have committed to specific “Safety by Design” principles to protect child safety.11

Additional Resources

Public service announcement on AI-generated CSAM from the FBI.
This public service announcement from the FBI reiterates that AI-generated CSAM is strictly illegal.

“Artificial Intelligence and Combatting Online Child Sexual Exploitation and Abuse” from the Department of Homeland Security.
This flyer from the DHS highlights the illegality of AI-generated CSAM and the opportunities for law enforcement to use AI to help identify victims and catch offenders.

TAKE IT DOWN Act from the US Congress.
This law prohibits exploitation from AI-generated deepfakes and requires platforms to take the content down.

“Safety by Design: One year of progress protecting children in the age of AI” from Thorn.
This article discusses the work by Thorn to help AI companies operate with “Safety by Design” principles that protect children.

OpenAI safety policy: Explicitly prohibits AI CSAM.

Meta safety policy: Explicitly prohibits AI CSAM.

Google safety policy: Explicitly prohibits AI CSAM.

Anthropic safety policy: Explicitly prohibits AI CSAM.

xAI safety policy: Explicitly prohibits AI CSAM.

No evidence of widespread AI-induced mental illness.

• Reports of “AI psychosis” or pathological attachment are isolated

• Cases typically involve pre-existing mental health conditions or substance use

• No epidemiological evidence links everyday AI use to population-level decline

Most AI interactions are neutral or positive.

• Millions use AI daily for productivity, learning, creativity, and casual support

• The vast majority of interactions show no harmful psychological effects

• AI is generally integrated as a helpful tool, not a destabilizing influence

AI chatbots show benefits for mental health support.

• 2025 studies find reductions in depression, anxiety, and distress

• Effect sizes are small-to-moderate and comparable to adjunct therapies

• Particularly valuable for accessibility and low-cost support

Broader youth mental health trends are not primarily driven by tech.

• Claims tying AI or social media to a teen mental health “epidemic” show weak correlations

• Observed effects are small and inconsistent across studies

• Broader drivers include economic stress, global disruption, and reduced autonomy

Rare problematic cases involve contextual vulnerabilities.

• Severe distress cases are linked to underlying mental health or life stressors

• AI is rarely the sole or primary causal factor

• Context matters more than technology exposure alone

Resources

• JMIR Mental Health – Effectiveness of AI Chatbots (2025)

• Frontiers in Psychiatry – AI Chatbots for Mental Health & Well-Being (2025)

• Dartmouth Study – First Therapy Chatbot Trial (2025)

• Pew Research Center – Teens, Social Media and Mental Health (2025)

• Nature Review – Social Media and Teenage Mental Health Epidemic? (2024)

What are data centers?

• Data centers are physical facilities that provide the infrastructure for data storage, compute, and inference.1

• Large data centers have been around for decades, powering the era of cloud infrastructure—AI is just a portion of data center usage.2

• With the rise of AI, companies are building out more data centers to train models and power user queries.

How much water do data centers really use?

• Data centers can use millions of gallons of water per day, put in perspective:

  • The Colossus 2 data center uses the water equivalent of two and a half In-N-Out joints.3

  • One pair of jeans uses the water equivalent of ~5.4M AI prompts.4

• Many data centers use non-water cooling systems and are developing more efficient methods.

• The major hyperscalers—Amazon, Google, Microsoft, and Meta—all plan to be water positive by 2030.5

Are data centers increasing electricity costs?

• Data centers account for approximately 1.5% of global electricity consumption (2024).6

• When data centers enter a new market, massive investments in new energy supply reduces overall electricity costs for consumers.7

  • Virginia, “the data center capitol of the world,” saw only a 3% rise in electricity rates compared to the 6.5% national average.8

  • North Dakota has had a 40% increase in electricity demand due to new data centers, yet their inflation-adjusted electricity prices have fallen.9

• Green energy mandates are a leading culprit for rising electricity prices, as states with stricter renewable energy policies have higher electricity rates.10

How do data centers affect the economy?

• Data center investment exceeds $50B annually in the U.S.11

• Between 2017 and 2023, direct employment through data centers grew five times more quickly than general employment in the U.S.12

• Over that same period, data centers have contributed $209 to $404 billion in annual labor income to American workers13

Charts

Additional Resources

“What is a Data Center?” from Amazon.
Overview of what a data center is from Amazon.

“There’s a reason electricity prices have been rising. And it’s not data centers.” from The Washington Post.
The article from The Washington Post shows that data centers are not the culprit for rising electricity prices.

“Energy and AI” from the International Energy Agency.
Report on AI data center energy use across the world.

“From Tokens to Burgers: A Water Footprint Face-Off” from Semianalysis.
A study from Semianalysis found that the xAI Colossus 2 data center uses the water equivalent of two and a half In-n-Out joints.

Build American AI tweet on AI data center usage.
One piece of paper uses the same amount of water as 2,550 prompts, one pot of water uses the equivalent of 5,000 prompts, and one pair of jeans uses the equivalent of 5.4 million prompts.

“Virginia’s electricity costs rise amid data center boom” from Axios.
The chart from Axios shows how Virginia, with some of the most data centers, had one of the smallest increases in electricity prices.

Decreasing property tax rates in Loudon County, Virginia from John Arnold.
In Loudon County, Virginia, the data center buildout has helped the county decrease property tax rates. The average savings for homeowners is about $3,400 per year.

”The Rise of AI: A Reality Check on Energy and Economic Impacts” from the National Center for Energy Analytics.
This thorough report from the National Center for Energy Analytics discusses the energy needs and potential benefits of data centers.

“Economic Contributions of Data Centers in the United States” from PWC.
A report from PWC highlights the economic benefits from data centers in the US.

What is AGI?

• Artificial General Intelligence (AGI) refers to a theoretical concept of an AI that exhibits general intelligence and capability equal to or beyond that of a human.1

• There is no consensus definition for what exactly AGI is or how to measure it.

• The term is frequently invoked when discussing dystopian, terminator-like AI.

Has AI gone rogue?

• Leading AI researchers have not seen instances of AI behaving unethically in the wild.2

• Instances of AI disobeying explicit commands or choosing unethical decisions arise in limited research scenarios where the AI is forced into logical paradoxes.3

Has AI blackmailed people to save itself?

• A study from Anthropic purported to find that AI would blackmail an employee to save itself from being shut down.

• In these studies, researchers tried hundreds of prompts to find one that led AI to blackmail when given no other option to complete its mission.

• Anthropic’s own researchers stated that they have never seen AI behave like this unless explicitly prompted to do so.4

Why do some AI companies talk about alignment?

• Some safety work is prompted by the existing patchwork of state-level regulations.5

• Many labs have dedicated safety teams ensuring that their AI models are safe and prohibit illicit outputs.

• Much of the research remains speculative, especially since AI models have not exhibited concerning behavior outside of testing conditions.