Are hallucinations by AI worse than hallucinations by humans?
https://www.bespacific.com/artificial-intelligence-and-constitutional-interpretation/
Artificial Intelligence and Constitutional Interpretation
Coan, Andrew and Surden, Harry, Artificial Intelligence and Constitutional Interpretation (November 12, 2024). Arizona Legal Studies Discussion Paper No. 24-30, U of Colorado Law Legal Studies Research Paper No. 24-39, Available at SSRN: https://ssrn.com/abstract=5018779 or http://dx.doi.org/10.2139/ssrn.5018779
This Article examines the potential use of large language models (LLMs) like ChatGPT in constitutional interpretation. LLMs are extremely powerful tools, with significant potential to improve the quality and efficiency of constitutional analysis. But their outputs are highly sensitive to variations in prompts and counterarguments, illustrating the importance of human framing choices. As a result, using LLMs for constitutional interpretation implicates substantially the same theoretical issues that confront human interpreters. Two key implications emerge: First, it is crucial to attend carefully to particular use cases and institutional contexts. Relatedly, judges and lawyers must develop “AI literacy” to use LLMs responsibly. Second, there is no avoiding the burdens of judgment. For any given task, LLMs may be better or worse than humans, but the choice of whether and how to use them is itself a judgment requiring normative justification.
An old complaint that has been solved by most organizations...
https://www.theregister.com/2024/11/20/data_is_the_new_uranium/
Data is the new uranium – incredibly powerful and amazingly dangerous
CISOs are quietly wishing they had less data, because the cost of management sometimes exceeds its value
I recently got to play a 'fly on the wall' at a roundtable of chief information security officers. Beyond the expected griping and moaning about funding shortfalls and always-too-gullible users, I began to hear a new note: data has become a problem.
A generation ago we had hardly any data at all. In 2003 I took a tour of a new all-digital 'library' – the Australian Centre for the Moving Image (ACMI) – and marveled at its single petabyte of online storage. I'd never seen so much, and it pointed toward a future where we would all have all the storage capacity we ever needed.
That day arrived not many years later when Amazon's S3 quickly made scale a non-issue. Today, plenty of enterprises manage multiple petabytes of storage and we think nothing about moving a terabyte across the network or generating a few gigabytes of new media during a working day. Data is so common it has become nearly invisible.
Unless you're a CISO. For them, more data means more problems, because it's stored in so many systems. Most security execs know they have pools of data all over the place, and that marketing departments have built massive data-gathering and analytics engines into all customer-facing systems, and acquire more data every day.
Keep America stupid? Why not learn to use the new tools?
https://www.bostonglobe.com/2024/11/15/opinion/ai-classroom-teaching-writing/
AI in the classroom could spare educators from having to teach writing
Of all the skills I teach my high school students, I’ve always thought writing was the most important — essential to their future academic success, useful in any profession. I’m no longer so sure.
Thanks to AI, writing’s place in the curriculum today is like that of arithmetic at the dawn of cheap and widely available calculators. The skills we currently think are essential — spelling, punctuation, subject-predicate agreement — may soon become superfluous, and schools will have to adapt.
… But writing takes a lot of time to do well, and time is the most precious resource in education. Longer writing assignments, like essays or research papers, may no longer be the best use of it. In the workplace, it is becoming increasingly common for AI to write the first draft of any long-form document. More than half of professional workers used AI on the job in 2023, according to one study, and of those who used AI, 68 percent were using it to draft written content. Refining AI’s draft — making sure it conveys what is intended — becomes the real work. From a business perspective, this is an efficient division of labor: Humans come up with the question, AI answers it, and humans polish the AI output.
In schools, the same process is called cheating.
(Related)
OpenAI releases a teacher’s guide to ChatGPT, but some educators are skeptical
OpenAI envisions teachers using its AI-powered tools to create lesson plans and interactive tutorials for students. But some educators are wary of the technology — and its potential to go awry.
Today, OpenAI released a free online course designed to help K-12 teachers learn how to bring ChatGPT, the company’s AI chatbot platform, into their classrooms. Created in collaboration with the nonprofit organization Common Sense Media, with which OpenAI has an active partnership, the one-hour, nine-module program covers the basics of AI and its pedagogical applications.
No comments:
Post a Comment