context: Tangermann, Victor. ( Feb 16, 2023). Microsoft: It’s Your Fault Our AI Is Going Insane They’re not entirely wrong. IN: FUTURISM (online). Last retrieved on 23 February 2023 from https://futurism.com/microsoft-your-fault-ai-going-insane
LLM types of technology and their spin-offs or augmentations, are made accessible in a different context then technologies for which operation requires regulation, training, (re)certification and controlled access.
If the end-user holds the (main) weight of duty-of-care, then such training, certification, regulation and limited access should be put into place. Do we have that, and more importantly: do we really want that?
If we do want that, then how would that be formulated, be implemented and be prosecuted? (Think: present-day technologies such as online proctoring, keystroke recording spyware, Pegasus spyware, Foucault’s Panopticon or the more contextually-pungent “1984”)
If the end-user is not holding that weight and the manufacturer is, and/or if training, (re)certification, access and user-relatable laws, which could define the “dos-and-don’ts,” are not readily available then… Where is the duty-of-care?
Put this question of (shared) duty-of-care in light of critical analysis and of this company supposedly already knowing in November 2022 of these issues, then again… Where is the duty-of-care? (Ref: https://garymarcus.substack.com/p/what-did-they-know-and-when-did-they?r=drb4o )
Thirdly, put these points then in context of disinformation vs information when e.g. comparing statistics as used by a LLM-based product vs the deliverables to the public by initiatives such as http://gapminder.org or http://ourworldindata.org or http://thedeep.io to highlight but three instances of a different systematized and methodological approach to the end-user (one can agree or disagree with these; that is another topic).
So, here are 2 systems which are both applying statistics. 1 system aims at reducing our ignorance vs the other at…increasing ignorance (for “entertainment” purposes… sure.)? The latter has serious financial backing, the 1st has…?
Do we as a social collective and market-builders then have our priorities straight? Knowledge is no longer power. Knowledge is submission to “dis-“ packaged as democratized, auto-generating entertainment.
Epilogue-1:
Questionably “generating” (see above “auto-generating entertainment”) —while arguably standing on the shoulders of others—rather: mimicry, recycling, or verbatim copying without corroboration, reference, ode nor attribution. Or, “stochastic parroting” as offered by Prof. Dr. Emily M. Bender , Dr. Timnit Gebru et al. is relevant here as well. Thank you Dr. Walid Saba for reminding us. (This and they are perhaps suggesting a fourth dimension in lacking duty-of-care).
Epilogue-2:
to make a case: I ran an inquiry through ChatGPT requesting a list of books on abuses with statistics and about 50% of the titles did not seem to exist, or are so obscure that no human search could easily reveal them. In addition a few obvious titles were not offered. I tried to clean it up and add to it here below.
bibliography:
Baker, L. (2017). Truth, Lies & Statistics: How to Lie with Statistics.
Barker, H. (2020). Lying Numbers: How Maths & Statistics Are Twisted & Abused.
Best, J. (2001). Damned Lies and Statistics: Untangling Numbers from the Media, Politicians, and Activists. Berkeley, CA: University of California Press.
Best, J. (2004). More Damned Lies and Statistics.
Dilnot, A. (2007). The Tiger That Isn’t.
Ellenberg, J. (2014). How Not to Be Wrong.
Gelman, A., & Nolan, D. (2002). Teaching Statistics: A Bag of Tricks. New York, NY: Oxford University Press.
Huff, D. (1954). How to Lie with Statistics. New York, NY.: W. W. Norton & Company.
Levitin, D. (2016). A Field Guide to Lies: Critical Thinking in the Information Age. Dutton.
O’Neil, C. (2016). Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. New York, NY Crown.
Rosling, H., Rosling Ronnlund, A. (2018). Factfulness: Ten Reasons We’re Wrong About the World–and Why Things Are Better Than You Think. Flatiron Books; Later prt. edition
Seethaler, S. (2009). Lies, Damned Lies, and Science: How to Sort Through the Noise Around Global Warming, the Latest Health Claims, and Other Scientific Controversies. Upper Saddle River, NJ: FT Press.
Silver, IN. (2012). The Signal and the Noise: Why So Many Predictions Fail – but Some Don’t. New York, NY: Penguin Press.
Stephens-Davidowitz, S. (2017). Everybody Lies: Big Data, New Data, and What the Internet Can Tell Us About Who We Really Are.
Tufte, E. R. (1983). The Visual Display of Quantitative Information. Cheshire, CT: Graphics Press.
Wheeler, M. (1976). Lies, Damn Lies, and Statistics: The Manipulation of Public Opinion in America.
Ziliak, S. T., & McCloskey, D. N. (2008). The Cult of Statistical Significance: How the Standard Error Costs Us Jobs, Justice, and Lives. Ann Arbor, MI: University of Michigan Press.
References
this post was triggered by:
thank you Katja Rausch
and by:
thank you Marisa Tschopp
Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜 In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’21). Association for Computing Machinery, New York, NY, USA, 610–623. https://doi.org/10.1145/3442188.3445922
“Verbatim copying” in the above post’s epilogue was triggered by Dr. Walid Saba ‘s recent post on LinkedIn:
This blog post on LinkedIn