AI Issues Key Points

Generative AI (GenAI) - like Gemini, ChatGPT, Claude, etc. - is not going away. There are many concerns about AI, but the cat has been let out of the bag and it won't go back in. As educators and as individuals, the best thing we can do in this new era of Artificial Intelligence is to help promote a more AI-literate society. People who jump right into AI without having formal, explicitly taught AI-literacy lessons are likely to fall into many pitfalls of AI issues and ethics.

***The following content uses "AI" throughout, but it's been pointed out that "GenAI" might be a more accurate term as "AI" is far more broad and outside the scope of this content. I may change this in the future to constantly say "GenAI", but for now, just assume that all the references to AI are specifically referring to generative AI.

AI can do AMAZING things and has so much potential, and I don't want to dissuade its use, but when using AI, you should be very aware of the following:

• AI Training - how does AI "learn"?

• Human-AI Relationship - What dangers are there when we interact with AI?

• AI + Ethics, Equity, and Law - What educational concerns should we be aware of beyond those already mentioned?

• AI is trained on giant data sets - more than what you could read/retain hundreds or thousands of lifetimes. (Big Data)

• AI doesn't know anything. It's simply a statistical machine giving you the statistically best answer based on its data - the "Confident Guesser". (Knowledge vs Data; Statistics)

• AI can be wrong - and often is. (Hallucinations; Bias; Statistics)

• AI has opinions/values based on the data they are trained by and further tuned by the companies overseeing them. (AI Values)

• AI is biased towards/against everything - especially for dominant groups and against minority/marginalized groups. Bias, by its very nature, is discriminatory. (Garbage-in, Garbage out; Algorithmic Bias; Bias in Big Data; )

• AI produces very realistic images. Seeing is no longer believing. (AI Slop; Deep fakes)

• AI trained on human data mimics human data. AI trained on AI data (trained on AI data trained on AI....) creates less-and-less human-like data (Model Collapse)

• AI doesn't disclose how it comes to its conclusions. (Blackbox vs Transparency).

• AI has major impacts on the environment - Water, Power, Carbon. (Environmental Impact)

• AI is designed to cause us to forget that it's not human and to put down our guard rails. (Anthropomorphism; The Eliza Effect; Banal Deception)

• AI is designed to be biased towards you and may prioritize pandering to you over giving accurate info. (Confirmation Bias; AI Sycophancy)

• AI is not an authoritative source. An AI's response should not hold more precedence than an authoritative human in their zone of specialization (Doctor, teacher, judge, etc.). (Anthropomorphism; authority)

• AI will take every job it can do better than humans - the problem is identifying what it can do better and what "better" means. (Workforce takeover)

• AI is not a substitute for human interaction. It does not love. It does not care. It does not understand. (Trust; Reliance; Relationships)

• AI will fill any role we allow it to fill in our personal lives - such as friend, significant other, confidant, personal counselor, medical-advisor, tax-advisor, etc.

• AI can cause serious concerns of mental health in those who over-anthropomorphize or over-"authoritize" AI.

• AI, when used to bypass our learning & struggle, could limit our mental growth (Cognitive Offloading; Academic Integrity)

• AI access is not equal - those with more access will have more opportunities to succeed (Digital Divide; Ethics of Access; Socioeconomic Privilege)

• AI creates major concerns regarding education & PII (Personally Identifiable Information) because every interaction discloses new information about you.

  • AI collects everything, even the stuff you don't realize you're sharing. (Privacy; Meta Data; Interests)

    • ***Correction, AI doesn't actually collect everything, but it could, and you don't know what exact information it considers important enough to collect. Some AI companies put in safeguards to protect user data, but this should not be an expectation.

  • AI infers everything else, and can make an infinite number of inferences and data connections.(Inferences; Writing Voice; Mannerisms;)

  • AI builds profiles and uses those profiles to tailor itself more and more to us (Banal Deception; AI Sycophancy; Confirmation Bias; Hallucinations; Authority; Predictions; Filter Bubbles).

    • ***Important note: Training and profile building are not the same thing.

  • AI doesn't need you to tell it who you are to connect your conversation to an existing profile of you. (De-anonymizing)

  • AI companies that sign data privacy agreements (DPAs) still profile students, but they don't use that data for training AI or advertising to students, and the data is owned by the educational institutions, not the AI company.

• AI's deep fakes are getting harder and harder to identify as AI generated. (Deep Fakes; Erosion of Trust; Liar's Dividend)

• AI created content ownership is difficult to pin down - who owns an AI-generated image? The AI-Company, the Prompter, the people whose works the AI was trained by and mimicked? (Copyright & Intellectual Property)

The content above was written entirely by me, the only AI input coming from AI critiquing my logic/arguments. This feedback caused me to add the "Generative AI" disclaimer and the correction on AI collecting everything.

However, there was a 3rd suggested change that AI gave to me, and I thought the conversation interesting, so I'm including it below.

It's nice that I convinced AI to side with me... but it makes me wonder - did I really make a valid point, or is Gemini's sycophantic nature just kicking in? Is Gemini just further proving my point that "AI is designed to be biased towards you and may prioritize pandering to you over giving accurate info"?