The Problems and Perils of AI

Artificial intelligence may be the biggest thing happening in the world today -- bigger than climate change, bigger than wars, bigger than presidential elections. Bigger, because AI looks set to play a role in how we deal with each of those issues and many others.

We plan to devote the meeting of the Northport Cracker Barrel on May 18 to the promises and perils of AI as they apply to the world at large and to our life in one small town in the American Midwest.

So much has been written about AI that it is hard to choose a single item as a reading to focus the discussion, but here is one that we like, written by Washington Post columnist Danielle Allen: "The next level of AI is approaching. Our democracy isn't ready." (Note: this gift link will expire on May 13. You might want to print it or download it if you want to save it to read later.) Please, use the comment box at the end of this post to share links to your own favorite readings on AI.

But don't take our word for it that AI matters. Ask AI itself! Here are two good places to experiment with AI:

  • Go to the OpenAI website and click the link that says Try ChatGPT.
  • Open the Microsoft Edge browser on your Windows PC. Click the Bing search logo in the upper-righthand corner. That will open a sidebar on the right of your screen where you can ask questions to Bing's version of OpenAI.
Here's an example. I opened the subscription version of ChatGPT (a little more powerful than the free version) and asked it a simple question: How will AI make life in small-town America better, and how might it be a threat?

Here is the answer that ChatGPT gave me:

AI has the potential to significantly improve life in small-town America, but it also brings certain risks and challenges that need to be considered. Here are some ways AI can make life better and some potential threats it may pose: 

Better: 

Improved healthcare: AI can enable better access to healthcare services through telemedicine, AI diagnostics, and personalized medicine. This can be particularly helpful in small towns with limited healthcare facilities and staff.

 Enhanced education: AI-powered tools and personalized learning systems can help improve education quality in rural areas where resources might be scarce. Online courses and virtual classrooms can connect students with experts from around the world. 

Agriculture and farming: AI-powered systems can optimize crop yields, reduce waste, and minimize the use of harmful chemicals, which can benefit small-town farmers and promote sustainable agriculture. 

Economic development: AI can enable small businesses to compete more effectively with larger companies by automating tasks, improving logistics, and expanding market reach. This can stimulate economic growth in small towns. 

Infrastructure maintenance: AI can help monitor and manage local infrastructure more efficiently, such as energy grids, water systems, and transportation networks. This can lead to more reliable services and better resource allocation. 

Threats: 

Job displacement: AI-powered automation could lead to job losses in various sectors, including manufacturing, agriculture, and retail. Small-town communities may be particularly vulnerable as they often rely on a limited number of industries for employment. 

Digital divide: As AI becomes increasingly integrated into daily life, communities without access to high-speed internet and modern technology may be left behind. This could exacerbate existing inequalities in small towns. 

Loss of privacy: AI-powered surveillance systems and data collection could lead to potential privacy violations, as the technology becomes more prevalent in various aspects of life. 

Bias and discrimination: AI systems can inadvertently perpetuate and amplify existing biases if not properly designed and monitored. This could lead to unfair treatment or outcomes for individuals from certain demographics or social groups. 

Dependence on technology: As small towns increasingly rely on AI systems, there is a risk that they may become overly dependent on these technologies, which can lead to vulnerabilities in the event of system failures or cyber-attacks. 

To maximize the benefits and minimize the risks of AI in small-town America, it is essential to invest in education, infrastructure, and digital equity, while addressing concerns related to job displacement, privacy, and potential biases.

Do you love AI or hate it? Find it confusing? Simplistic? Scarry? Overrated? Come give us your take on May 18! 


Comments

  1. AnonymousMay 2, 2023 at 7:59 AM

    This is a good topic for discussion. Articles are appearing daily on the benefits and concerns about AI. I've been using it for a month or so for Internet searching and have found it beneficial. I have no idea whether I've been given false information. There is an interesting article in the May 1 NYT on how AI chatbots hallucinate: https://www.nytimes.com/2023/05/01/business/ai-chatbots-hallucinatation.html?searchResultPosition=7

    Ed Ketterer

    ReplyDelete
  2. AnonymousMay 5, 2023 at 6:52 AM

    And here's another NYT article, this one about Geoffrey Hinton, the "godfather" of AI, who now has his doubts: https://www.nytimes.com/2023/05/01/technology/ai-google-chatbot-engineer-quits-hinton.html?searchResultPosition=1

    Greg Nobles

    ReplyDelete
  3. Gerald SchatzMay 15, 2023 at 3:11 PM

    See this Brookings analysis re AI and the 2024 elections:https://www.brookings.edu/blog/techtank/2023/05/03/how-ai-will-transform-the-2024-elections/

    ReplyDelete
  4. AnonymousMay 15, 2023 at 3:15 PM

    Gerald Schatz: Real thoughts about artificial intelligence

    As a history student, I learned only two things well: There's lots that we can't know, and beware of extravagant statements and claims.

    Reacting to China's new regulation of what has come to be called artificial intelligence (AI), a tech-savvy law professor colleague laughed, "Good luck, CCP," which will find it hard to keep up. The changes have come so fast. The major, general purpose search engines already have transformed. By anticipating the next word or phrase based on statistical analysis, they now often fail to respond precisely to precise queries but instead seem to regard all queries as plain language and then respond accordingly imprecisely, with a heavy advertising push.

    Within a few months we have seen creepy reports of seemingly introspective AI products, and we have seen scary reports that some of these products have scored remarkably well on medical and legal qualification exams. The scariest prospect that I've seen is implied by W. Stone Holden & T. Adams, "Hey Siri, Who's in Charge?" Marine Corps Gazette, May 2023, pp. 65-68.

    AI has piqued a lot of public conversation among medical and legal ethics worriers, at the same time prompting technological optimism in both medicine and law. The concerns include:

    -- Authenticity. Is the text from a person or is it machine-generated? Are the authorship and sources verifiable and verified? Not entirely a new problem. Think personally addressed pitches for money. Think cut-and-paste term papers. Think boilerplate documents. Think forgery.

    -- Accuracy: Is what is presented as fact actually factual? Biases and assumptions underlying the answer? How reliable are the contents of the source library? Translation error? Transcription error? Selection error? Uncritical reiteration and reinforcement? Garbage in, garbage out.

    -- Adequacy: How responsive is the answer?

    -- Dirty tricks: On a grand scale, this is the disinformation and fake-news problem.

    -- Provenance, transparency, timing, honesty: What's the basis for the answer? Sources? Indicia of credibility of sources and sufficiency of search? How current is the information underlying the answer? Some testers report AI programs' yielding make-up footnotes--correct in form and otherwise fictional.

    -- Confidentiality: AI programs train on the questions put to them, and they seek and recognize patterns.

    -- Utility: This is the great temptation. It may be a marvelous time-saver when the answers can't hurt. Already there are assertions in law practice and medical practice that the time saved with AI will allow more clients and patients to be helped at less cost.

    -- Misuse: AI is no substitute for professional responsibility and professional judgment. Human beings are individuals, each different, not statistical summations.

    Irony isn't quite the word for it, but we are seeing a strange cultural reversal. For several years now, in many ways the Internet has functioned unidirectionally. Based on advertising targets and marketing algorithms, many web sites have talked at us but not with us. Marketers--whether of goods, ideas, or politics--have told us via web sites, emails, and texts that they were interested in our concerns; no, only in where we went and what we bought or might buy. The AI models might sin the same way. But they depend on our input--what the trade now terms prompts. Word on the street is that a group at MIT is developing a library of prompts. So we will need new vocabulary if our relationship with search engines, at least, is to be two-way, a seeming conversation. Our skeptical prompts will get some not necessarily authoritative response. But it will be two-way, as long as we stay skeptical.

    Consider Gertrude Stein's last words (quoted by Alice B. Toklas, in What Is Remembered):

    "What is the answer?"

    [Toklas was silent.]

    "In that case, what is the question?"








    ReplyDelete
  5. AnonymousMay 29, 2023 at 6:39 AM

    Here is an interesting article about a lawyer using AI. Don Frerichs
    https://www.nytimes.com/2023/05/27/nyregion/avianca-airline-lawsuit-chatgpt.html?campaign_id=9&emc=edit_nn_20230528&instance_id=93679&nl=the-morning&regi_id=64678603&segment_id=134120&te=1&user_id=ec7369f396a96c3cc70a84e3843a39c7

    ReplyDelete

Post a Comment

Popular Posts