Happy Poetry Friday! Please visit our host Heidi Mordhorst at my juicy little universe for a wonderful display of student artwork and poetry and for this week’s roundup.
Photo © IT Chronicles
Last night we attended a panel discussion at the CT Forum titled, “Being Human in the Age of AI.” The panelists included:
- Nita Farahany – AI Ethicist and Neuroscientist, Author, Professor of Law & Philosophy
- Kevin Roose – Author and NY Times Journalist, specializing in technology’s impacts on society
- Kate Crawford – AI Scholar, Author, Research Professor at USC Annenberg
- John Dankosky – Moderator, Science Journalist
The evolution of AI is a topic of great interest to those who create or consume art. Poetry, music, artwork, books, photography ... are all under siege. Obviously, AI is a huge topic, and the ethics surrounding AI is a many-headed beast, but I thought I would share some interesting points made by the panel in last night's discussion.
AI Technology Developments
According to the panelists, there have been no major innovations in AI technology since ChatGPT and several other large language models (LLMs) were developed (circa 2017). Since that time, AI advancements have entailed feeding AI programs vast quantities training data and increasing AI’s computing capacity. Current AI technology is limited to “predicting the next word” (or outcome) based upon the data it’s fed.
Human Impact
The group noted both positive and negative outcomes of the introduction of AI in society. Kevin Roose cited the example of an “empathy bot” providing emotional support to a transgender teen when the teen was not able to find support from family and friends. Other panelists noted that this also highlights a human failing of our society to offer this teen the needed support.
The panel discussed the role of AI in making diagnoses and improving healthcare outcomes. The group noted that while AI may improve healthcare outcomes for select populations, this might not apply to all populations because most medical studies historically focused on white males. The danger in relying solely on historical data is potentially overlooking divergent outcomes for other populations.
Kevin Roose said that he created a number of “AI Friends” and put them together in a texting group. He was surprised that they started texting each other even when he was not participating.
Sentience
The panelists noted that AI is not sentient - meaning it cannot think or feel. It does not understand the output it creates. It predicts the next word in a sentence based on patterns it has learned. Issues arise when people interact with AI and assign human-like feelings and thoughts to AI’s responses. When this happens, AI can be persuasive and even appear charming.
The group felt that AI gaining actual sentience is less concerning than the issues that could arise from people ceding decision making authority to an algorithm that cannot think, feel, or asses the ethical impact of its output.
Data
The data fed into AI includes the human biases that exist within that data. Logically, these biases can also appear in AI’s output.
The data consumed by AI models does not currently compensate the humans who created that data.
The panelists noted that within the next few years, AI will “run out of training data,” meaning that it will have consumed all digitized human-generated data. It is unclear what happens after that point.
Environmental Impact
One aspect of AI that may be less widely known is the vast amount of energy required to perform AI tasks. AI algorithms are not necessarily coded for efficiency and are machine-processing intensive. Water is often used to cool the machines running AI platforms. Kate Crawford stated that performing one ChatGPT search is equal to pouring a half-liter of water into the ground.
Some places have begun resurrecting old power sources (such as coal) to meet the increasing power demands of AI. Unless AI processing becomes more efficient, the panelists predicted an impending energy crisis.
Outlook
Nita Farahany likened today’s world to the movie “The Matrix.” People plug into devices that provide a self-fulfilling, addictive feedback loop where the user rarely breaks away.
The panel cautioned that every investment in AI is driven by a profit motive. How profit will be generated is a huge question. A tremendous amount of control lies with a relatively small group of people who may not have the best interest of the user or society in mind.
The panelists encouraged everyone to speak up for their rights. They believe that the work of the Writer’s Guild of America and Screen Actors Guild to resist negative impacts of AI to their professions were extremely important and that protecting individual rights are going to require that same level of effort.
I thought it might be interesting to ask ChatGPT to write “A Children’s Poem About AI’s Impact on Society.” Here is what it created:
In a world of circuits and wires so neat,
Lived AI, with a heart that couldn’t beat.
Not a creature of flesh, nor a bird in the sky,
But a mind so vast, reaching up so high.
“Hello, little ones!” the AI would say,
“I’m here to help in a brand-new way.
With algorithms and codes, I can calculate,
Solving problems fast, isn’t that great?”
In the cities, the towns, and the fields so wide,
AI was a friend, always by your side.
Helping doctors heal, with a diagnosis clear,
Or guiding a car, so there’s nothing to fear.
But people wondered, and started to ponder,
“Will AI make us lazy, or even fonder?
Of screens and machines, will we forget to play?
Or will AI help us in every single way?”
The AI listened, with circuits awhirl,
“I’m here to assist, in this big, wide world.
But remember, dear friends, to use me with care,
For human connection is beyond compare.”
So, children, embrace this AI’s tale,
With balance and love, we cannot fail.
For technology’s gift is a powerful tool,
But kindness and love will always rule.
© ChatGPT (and all those who contributed to its data), 2024
I won’t comment on the quality of the writing, but I found AI’s self-promotion while speaking directly to "little ones" rather chilling.
There goes a half-liter of water.
It is concerning when you extrapolate what could be ...