Simplify USA Immigration Policies

Against Artificial Intelligence and Data Centers, As They Are Now

Every decision we make has to be a “net good”. How do we travel to/from work, errands? What do we eat? What do we wear? We make all of those choices, consciously or not, after weighing a variety of factors. Our choices are our conclusions that those options were better than other options.

The use of Artificial Intelligence (AI) must meet that same goal, that it’s a “net good”.

It’s not.

AI claims to

1. Do tasks better. 
1. But AI has made mistakes, sometimes serious ones. “An AI chatbot set up to help small firms quickly obtain advice on the legal obligations and regulations businesses have to adhere to in New York starts telling business owners to break the law.”
2. “Approximately 90% of AI startups fail within their first year of operation.” Too few ambitious entrepreneurs recognize that AI is simply a tool, no different from Microsoft Office, and any proposed business or project using it is subject to the same business success principles as those without that tool. Does its output fulfill a need not already satisfied? 
3. With these experiences, can NTIA really be sure that “Data centers are important enablers for economic growth….” One MIT Economics professor’s model disagrees. His model suggests modest growth at best, and a real risk of worsening income inequality for laborers displaced from jobs. 
2. Do tasks faster: for the amount of data validation and AI system retraining needed, is this really true? 
1. “Garbage In, Garbage Out” is just as valid for artificial intelligence as it is for any other programming and data analysis.
2. The data input to AI is never complete. If the Internet is the source, that alone is an incomplete data source. Much information remains in formats that can or will never be digitized. New observations and theories happen in labs and fieldwork - e.g. geology, archaeology, science, other - that may take years to be reduced to a format that can be “scraped” or “scanned” for any AI program. 
3. “….if AI companies are found liable to third parties for copyright infringement in the pending litigation …, the cost of using generative AI services may increase….” It’s not clear whether the current AI data collection efforts are legal.
3. Provide jobs
1. The state of Washington did an audit. This article concludes that the $millions in tax breaks far outweighed the job gains.
2. In Illinois, “The 21 data centers that executed MOUs between June 28, 2019, and December 31, 2023, and received exemption certificates, reported 429 new jobs, 419 of which are associated with the operations and maintenance of those data centers. The data centers reported $11,029,423,534 in capital investments and $465,332,331 in reported exemption value, or 4% of the total investment.” So each of the 429 jobs cost Illinois ~$1.08Million in tax revenue. 
3. In Nevada, between 2016 and 2022, 271 jobs created, ~$312Million in tax exemptions, therefore each job cost Nevada ~$1.151Million in tax revenue. 
4. Georgia’s auditors tried to estimate the relative tax benefit of the state’s technology tax exemptions and had difficulty. Its tentative conclusion published December 2022 wasn’t promising. “using the hypothetical scenario of an $800 million HTDC, forgone state revenue would total approximately $80 million. Economic activity attributable to the exemption would result in approximately $19.4 million in additional state taxes during construction. HTDC operations would result in another $250,000 to $275,000 in state taxes each year.” 
4. Be reliable 24/7/365 
1. “Fifty-five percent of organizations reported having experienced a data center outage in the past three years.” “According to Dun & Bradstreet, 59% of Fortune 500 companies endure a minimum of 1.6 hours of downtime each week, averaging a weekly cost ranging from $643,200 to $1,056,000.” 
2. DownDetector has a separate webpage of major online service providers, including OpenAI, that tracks outages almost in realtime. 
3. One underreported, huge risk to the safety and reliability of AI and data centers are the undersea cables that move data. “over 95 percent of international data and voice transfers are currently routed through the many fiber optic cables that crisscross the world’s sea floors.” Those cables are vulnerable to everything from sharks to nature to human attacks. Availability and improvements to AI depend on this data. 
1. “Citing unnamed sources in the State Department, the Wall Street Journal reported in May that U.S. officials privately warned tech firms including Google and Meta that undersea cables in the Pacific region could be vulnerable to spying by Chinese repair ships.” 
2. Red Sea cables have been threatened because of the Mideast war, and African cables were cut in February and May, 2024. 
3. In January, 2022, “The strength of the Hunga-Tonga-Hunga-Ha’apai eruption severed internet connectivity with Tonga, causing a communication blackout at just the moment that a crisis was unfolding.” 
4. Every time there’s a technology advance, malicious actors exploit it. The more sophisticated the technology, the more difficult it is to mitigate and track these criminals. Cybersecurity hygiene methods become more difficult as AI methods make malicious communications more difficult to detect. 
1. The CyberManagement Alliance publishes a blog of recent cyberattacks monthly. The “list of cyber attacks, ransomware attacks and data breaches [cover a] diverse list of organisations that were compromised this month point to one resounding truth all over again - nobody is immune to the wrath of malicious threat actors on the internet.” 
5. Cryptocurrency miners were early creators and users of data centers. Its growth is out of control and causing problems. The underlying blockchain systems that define the decentralization basic to define and verify cryptocurrency transactions require banks of servers. “The University of Cambridge's Bitcoin Electricity Consumption Index estimated worldwide bitcoin mining used 121.13 terawatt-hours of electricity in 2023," more than the equivalent of the entire electricity consumption of the Netherlands. 
1. There are now more than 13,000 different cryptocurrencies and 724 exchanges. No one guarantees that the original Bitcoin principles outlined by the mysterious Satoshi will be kept. “The astonishing growth rate isn't entirely good news -- or truly representative of the crypto market. Many new cryptocurrencies have little purpose other than making money for their developers. The total number of cryptocurrencies also includes lots of dead coins, including abandoned projects and scams.” 
2. In 2023, worldwide cryptocurrency fraud amounted to $5.587USD Billion, $4.9USD Billion in the USA. 
3. Interestingly, cryptocurrency mine owners have begun leasing their capacity to AI. 

Many of these issues parallel concerns raised in a 2017 Pew Research article as the Internet of Things exploded into our lives. Several experts suggested the following themes:

Theme 1: People crave connection and convenience, and a tech-linked world serves both goals well
Theme 2: Unplugging isn’t easy now, and by 2026 it will be even tougher
Theme 3: Risk is part of life. The Internet of Things will be accepted, despite dangers, because most people believe the worst-case scenario would never happen to them
Theme 4: More people will be connected and more will withdraw or refuse to participate
Theme 5: Human ingenuity and risk-mitigation strategies will make the Internet of Things safer
Theme 6: Notable numbers will disconnect
Theme 7: Whether or not people disconnect, the dangers are real. Security and civil liberties issues are being magnified by the rapid rise of the Internet of Things. 

The desire to disconnect has grown, especially among older generations. Approximately 50% of  “smart” appliances purchasers don’t connect them to the Internet. Ten months ago (~December 2023), a car owner received many comments to his desire for disconnected cars; there’s no data available of how many share his opinion. The Global Day of Unplugging, generally the first weekend of March annually, began in 2009, as more people recognized the health benefits of disconnecting from all online activity.

If the alleged benefits of AI can’t be proven, then what about its costs?

1. This 2024 United Nations report on the Digital Economy concludes that the increasing demand for the latest technology, including AI, damages the environment and increases income inequality worldwide. It includes the entire lifecycle of the latest hardware, such as data center equipment, as well as what it takes to gather and transmit the data. “Developing countries bear the brunt of the environmental costs of digitalization while reaping fewer benefits. They export low value-added raw materials and import high value-added devices, along with increasing digital waste. Geopolitical tensions over critical minerals, abundant in many of these countries, complicate the challenges.”
1. “From 2018 to 2022, electricity consumption by 13 of the largest data centre operators more than doubled. Worldwide, data centres are estimated to have consumed as much energy as France in 2022 – 460 terawatt-hours (TWh) of electricity. Their energy consumption is expected by the International Energy Agency to double to 1,000 TWh in 2026.

Such consumption can strain local electricity grids. For example, data centres in Singapore accounted for around 7% of the country’s electricity demand in 2020, and in Ireland that share was as high as 18% in 2022.

Cryptocurrency technologies are also energy intensive. For example, Bitcoin mining's global energy consumption rose 34 times between 2015 and 2023, reaching an estimated 121 TWh.

Digitalization’s water consumption is also growing, which is cause for concern in a world where two billion people still lack access to safe drinking water.

In 2022, Google’s data centers and offices consumed more than 21 million cubic meters of water. Newer technologies, such as generative AI, also require more potable water for cooling servers.

In the United States, one-fifth of data centre servers’ direct water footprint reportedly comes from watersheds that are moderately to highly water-stressed.”

2. According to the USDA 2022 farmland survey, the number of acres considered farmland dropped from 900 million acres in 2017 to 880 million acres in 2022, a 2.2% drop. While many factors caused that, especially residential development, it’s noteworthy that the newest data center projects now require >100 acres each, and that forces them into rural areas. 
1. Virginia’s farmland pays the price. 
2. Maryland is considering construction of 70 miles of transmission lines that would affect 4,000 acres in 3 counties. 
3. Texas has at least 20,000 acres devoted to existing and planned data centers. 
4. Georgia has at least 1,600 acres in data centers, existing and planned.

We have a finite amount of land, and 8 billion people who need to eat, over 330 million in the USA. Once that land is developed it never returns to its original state. 

1. We have a finite amount of water, especially potable or drinkable water. Our lives, and the lives of all living things, depend on it. 

The water consumption of data centers can be illustrated with two examples: i) hyperscale data centers and ii) wholesale and retail data centers. In both cases, a single data center uses several million gallons/liters of water each year.

Water Consumption	Hyperscale	Wholesale and Retail*
per day	550,000 gallons (2.1 million liters)	18,000 gallons (68,000 liters)
per year	200 million gallons (760 million liters)	6.57 million gallons (24.9 million liters)

1. “Much of the USA is at risk of not having enough water for drinking or farming or any other purpose.”  “…water resources are withdrawn at a faster rate than they are replenished.”  

2. Noise pollution. “Excessive noise has been linked to hearing loss, tinnitus, sleep deprivation, cardiovascular disturbances, mental health impairment, impaired task performance, aggressive behavior, and chronic fatigue. Noise levels throughout the world are growing at alarming levels.”  
1. Inside a data center hundreds of servers operating in a small space can create noise levels of up to 96 db(A), enough to cause hearing loss in people who endure the noise levels for an extended period.
2. Outside of data centers, there’s been limited research of the effects of the low-frequency noise experienced by those living near them. But “Machine-generated Infrasound (<20 Hz) and low frequency noise (<200 Hz) (ILFN) are physical agents of disease that have gone unregulated by governmental agencies in Western Europe, North America, and Oceania, despite their proven pathological action on biological systems. ILFN has been operating (for decades now) as an unrecognized confounding factor in many pathologies common in modern societies, such as cardiovascular disease, respiratory illnesses, non-specific neuropathology, and some forms of cancer.” 
1. In the USA “Some associations were observed between exposure to LFN and annoyance, sleep-related problems, concentration difficulties and headache in the adult population living in the vicinity of a range of LFN sources.”
2. Animals and plants hear or sense noise in ranges different from, sometimes greater than, humans. All react adversely, sometimes severely, to increased noise.  
3. Exterior lighting needed for data centers both during and after construction has disturbed nearby residents. Data centers generally use LED lighting. LED light pollution “interferes with predator/prey relationships and plant phenology; disorients migrating birds and newly hatched sea turtles, among other species; and attracts insects to artificial lights so they more easily become prey to birds, and more. …Excessive artificial light can also interfere with the symbiosis that is critical to plant pollination and reproduction, and can negatively impact amphibians’ foraging and reproduction habits.” 

3. “Exposure to artificial outdoor light at night is associated with numerous detrimental health effects including sleep disruption, obesity, depression, anxiety, memory dysfunction, atherosclerosis, and cancer.” There may even be a correlation between the amount of nighttime outdoor lighting and Alzheimer’s disease, especially for those under 65 years old. 

2. AI curtails our thinking and reasoning processes.
1. The same MIT professor referenced earlier has other research suggesting consumer harm related to AI. “….the vast amounts of data collected by online platforms may also enable behavioral manipulation, harming users….” 
2. The latest AI chatbots encourage relationships with them rather than with other people.  
3. And AI may self-destruct. “….training generative AI models on AI-generated content causes models to erode. In short, training on AI content causes a flattening cycle similar to inbreeding….by just the fourth iteration, the model was spouting complete gibberish: ‘To cook a turkey for Thanksgiving, you need to know what you are going to do with your life if you don 't know what you are going to do with your life if you don 't know what you are going to do with your life…’ "
1. If that Habsburg, or cannibalistic, effect occurs, what happens then to the data centers and equipment within?

Challenge your assumptions that “Data centers are important enablers for economic growth and technological development. Their capabilities for data processing, ubiquitous connectivity, secure storage, cost-efficiency, and economy-wide job creation, among others, yield substantial benefits.” They don’t. They cause more harm than good to our environment, they challenge our cybersecurity capabilities, and their alleged economic benefits aren’t proven. Their primary use, Artificial Intelligence, can cause more harm than good to our ability to think and reason.

The USA signed the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law early in September, 2024.  As you review the many responses, ensure that all resulting recommendations abide by it, especially these sections: 

Article 8 – Transparency and oversight Each Party shall adopt or maintain measures to ensure that adequate transparency and oversight requirements tailored to the specific contexts and risks are in place in respect of activities within the lifecycle of artificial intelligence systems, including with regard to the identification of content generated by artificial intelligence systems. 

Article 9 – Accountability and responsibility Each Party shall adopt or maintain measures to ensure accountability and responsibility for adverse impacts on human rights, democracy and the rule of law resulting from activities within the lifecycle of artificial intelligence systems.

Article 14 – Remedies 1 Each Party shall, to the extent remedies are required by its international obligations and consistent with its domestic legal system, adopt or maintain measures to ensure the availability of accessible and effective remedies for violations of human rights resulting from the activities within the lifecycle of artificial intelligence systems.

Any regulations and laws are only as good as they are enforced.

That AI Convention cannot just be words.  As the 5,000+ data centers in the USA have been built, over and over again the same pattern emerged. Word leaked out that an “unnamed developer” had purchased dozens or hundreds of acres of land, no purpose mentioned. The related state’s utility company and state Public Service Commission (PSC) quietly proposes massive changes, usually increases, in its power transmission grid. Those proposed changes are approved with minimal public notice or opportunity for comment. Then the same utility company and PSC approves utility rate increases to pay for the new infrastructure. Only then does the “unnamed developer” request zoning changes and tax exemptions from the affected communities, promising inflated numbers of jobs and other local improvements. Environmental impacts aren’t mentioned or considered unless the local community protests vigorously. Local governments, starstruck by AI and accompanying promises, approve the data center - and never or rarely audit or otherwise ensure than any related promises of sustainability or economic improvements are kept.

This pattern must stop. NTIA’s recommendations should coordinate with the AI Convention in requiring transparency and accountability for all existing data centers. All should become or be compatible with sustainability recommendations. Sustainability and economic development promises must be subject to periodic audits to ensure that they’re kept, and monetary penalties and operational restrictions should be imposed if they are not.

References

1. https://tech.co/news/list-ai-failures-mistakes-errors 
2. https://www.cio.com/article/190888/5-famous-analytics-and-ai-disasters.html 
3. https://kitrum.com/blog/why-do-ai-startups-fail-5-lessons-learned-from-startup-failures/ 
4. https://www.theregister.com/2024/05/29/ai_gdp_inequality/
5. https://economics.mit.edu/sites/default/files/2024-08/The%20Simple%20Macroeconomics%20of%20AI.pdf 
6. https://shelf.io/blog/garbage-in-garbage-out-ai-implementation/ 
7. https://www.ropesgray.com/en/insights/alerts/2024/04/ai-and-the-copyright-liability-overhang-a-brief-summary-of-the-current-state-of-ai-related 
8. https://www.propublica.org/article/washington-data-centers-tech-jobs-tax-break 
9. https://www.seattletimes.com/seattle-news/times-watchdog/data-centers-got-a-huge-tax-break-to-create-wa-jobs/ 
10. https://dceo.illinois.gov/content/dam/soi/en/web/dceo/aboutdceo/reportsrequiredbystatute/2023-data-centers-annual-report.pdf
11. https://www.leg.state.nv.us/Division/Research/Documents/RTTL_NRS231.0685_2023.pdf
12. https://www.audits2.ga.gov/wp-content/uploads/2023/01/Data-Center-Sales-Tax-Exemption-Summary.pdf 
13. The full Georgia report, including a comprehensive literature review  https://www.audits.ga.gov/ReportSearch/download/29072 
14. https://www.datacenterknowledge.com/outages/top-data-center-outage-trends-and-strategies-for-reducing-risk
15. https://www.techradar.com/pro/the-true-cost-of-outages-and-why-monitoring-ai-dependencies-is-crucial
16. https://downdetector.com/online-services/
17. https://www.cnbc.com/2024/07/16/next-front-in-us-china-tech-battle-is-underwater-internet-cables.html  
18. https://developingtelecoms.com/telecom-technology/optical-fixed-networks/16697-east-africa-internet-disrupted-again-by-new-subsea-cable-cuts.html 
19. https://www.theguardian.com/environment/article/2024/aug/09/what-lies-beneath-the-growing-threat-to-the-hidden-network-of-cables-that-power-the-internet 
20. https://www.cm-alliance.com/cybersecurity-blog/august-2024-biggest-cyber-attacks-data-breaches-ransomware-attacks
21. https://www.cnet.com/home/energy-and-utilities/how-much-power-does-crypto-use-the-government-wants-to-know/ 
22. https://www.fool.com/investing/stock-market/market-sectors/financials/cryptocurrency-stocks/how-many-cryptocurrencies-are-there/ 
23. https://www.ic3.gov/Media/PDF/AnnualReport/2023_IC3CryptocurrencyReport.pdf 
24. https://time.com/6993603/ai-bitcoin-mining-artificial-intelligence-energy-use/ 
25. https://www.pewresearch.org/internet/2017/06/06/the-internet-of-things-connectivity-binge-what-are-the-implications/ 
26. https://news.ycombinator.com/item?id=38197133
27. https://unctad.org/publication/digital-economy-report-2024
28. https://capitalregionland.org/2024/03/virginia-losing-farmland-at-record-pace/
29. https://dgtlinfra.com/data-center-water-usage/
30. https://www.wri.org/applications/aqueduct/water-risk-atlas/#/?advanced=false&basemap=hydro&indicator=w_awr_def_tot_cat&lat=30&lng=-80&mapMode=view&month=1&opacity=0.5&ponderation=DEF&predefined=false&projection=absolute&scenario=optimistic&scope=baseline&threshold&timeScale=annual&year=baseline&zoom=3
31. https://watercenter.sas.upenn.edu/splash/water-stress-water-scarcity
32. https://noisefree.org/noise-health/
33. https://www.mdpi.com/journal/ejihpe/special_issues/ILFNPAPHE
34. https://pubmed.ncbi.nlm.nih.gov/26994804/
35. https://www.linkedin.com/pulse/noise-pollution-harm-animals-anumeenacare
36. https://new.nsf.gov/news/effects-noise-plants-persist-over-long-periods
37. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11412842/
38. https://economics.mit.edu/sites/default/files/2024-08/When%20Big%20Data%20Enables%20Behavioral%20Manipulation.pdf 
39. https://www.technologyreview.com/2024/08/12/1096202/how-people-actually-using-ai/ 
40. https://futurism.com/ai-slowly-killing-itself 
41. The AI Convention. https://rm.coe.int/1680afae3c