PUBLICATION ANNOUNCEMENT
Transferring Publication of Ethics & AI to Aman Dasgupta
Neglected publication moving to a younger, brighter, more energetic entrepreneur
OpenAI released ChatGPT in November 2022, and quickly became the biggest whale in the AI sea, making a huge splash. Other heavyweights have emerged and are vying for market share as governments grapple with regulation and safety.
One year later, It feels like a good time for me to step back from this neglected publication and let Aman oversee the many deep dives of discussion needed in ethics and AI.
Aman Dasgupta offers highly educated, multilingual talents.
I’ve worked with many people from India, Iran, China, Taiwan, Russia, and other parts of the world in tech and writing. Many competitive tech companies recognize the value of a global approach and seek the best of the best regardless of physical base.
Aman is fluent in 5 languages and attended college in the US. He can help writers appeal to a primarily English-speaking base of readers and is more familiar with regional references than I am.
I started Ethics & AI at the end of 2022 but didn’t dedicate much time or attention to attracting writers or promoting the publication.
My employer designs and builds state-of-the-art technology including AI tools. Besides a long tech career, I have other interests and responsibilities as well. Writing and editing under my own profile and my snarlcastic alter-ego Lizzie Lizard Brain takes all my spare time.
General ethical concerns
The proliferation of AI systems and AI-generated writing, music, video, and art are concerns to creatives, primarily due to ownership and distribution rights, privacy, and safety.
Many people are unaware everything they post on public websites, social media platforms, or in an app is vulnerable to web scraping. Their data ends up in a repository like Common Crawl, which is then used to train AI Large Language Models (LLMs).
From my piece on challenging hate speech and misinformation —
As many discovered, new AI Large Language Models (LLMs) are trained on repositories of posts retrieved by web crawlers, so even information that was removed from websites may still be included in AI recaps of a person’s career or past.
Examples include false claims a professor was investigated for a sexual harassment claim, and a white hat hacker was responsible for a notorious virus. Both stories are false and were removed from the sites where they originally appeared, but were incorporated into AI training material scraped from Common Crawl.
The individuals falsely portrayed as guilty parties have been unsuccessful in removing slanderous misinformation from the OpenAI and Google’s Artificial Intelligence (AI) LLMs.
This is going to be a privacy and security concern as kids become adults and their once-private Facebook, Twitter, and other posted information becomes available as quick AI responses.
In addition to previous false claims, individuals have discovered AI is used to generate online porn with their images in the form of photos and videos. This has become a problem for 14-year-olds who are still developing a comfort level with their sexuality.
As people find it more difficult to discern fact from fiction, there are AI agents like “Ask Jesus” and similar apps to guide them with the owner’s slant on morality, obedience, and religious interpretation of ancient and modern writings.
When people use AI tools like ChatGPT, Bard, Jasper, or others, they pick up wrong information, repeat it, and the errors are fed back into AI.
Following the adage “A lie spreads around the world before the truth gets its boots on,” sensational announcements, opinions, and propaganda are used to influence public opinion. This was the case when Hamas terrorized Israel in October 2023 and deepfake images of mutilated babies were shared.
Some social media and tech platforms are taking steps to control the spread of misleading content, such as YouTube’s new policy requiring disclosure of AI-generated content. Others like Twitter (X) promote Infowars campaign material from China, Russia, and other adversaries.
Recycling misinformation creates more bad answers and the tools become less reliable.
Accuracy has gotten worse in every area of AI processing where it once made errors and those errors were then recycled into training, including basic math functions, statistical analysis, and some coding.
Relying on AI for most answers is like eating sugar for every meal — there needs to be a balance of nutrients from other foods or we become insulin-dependent. Reliable, contextual source material is needed to counter AI.
The AI-generated image I used above is interesting, but the real thing has a heart. So does Wyland. He works on oceanic conservation and donates time and talent — which he can’t do if he doesn’t make a living.
AI detection
Artists report losses in revenue and opportunities at the same time developers work to watermark or otherwise identify AI-generated art.
Artificial Intelligence is used to power reliable text AI-detection tools such as CopyLeaks, GPTZero, and OriginalityAI. These tools have the added benefit of identifying plagiarism in published works.
InVid and Google’s Lens are popular tools for identifying deepfake videos and photos.
When ChatGPT first made waves I joined the fray of fascinated users and tested the limits of many of the tools. I used some AI-generated art and now await the results of several pending lawsuits that could require this content to be removed.
The same could happen with text generated in the style of others.
AI bias and regional/cultural discrimination
AI can and will pick up on discriminatory practices ingrained in the datasets it is trained on. Remember when Amazon developed (and then scrapped) an AI tool for recruiting? The aim was to automate the search for talent, but since it used parameters to find candidates who resembled Amazon’s previous hires, the tool quickly learned to prefer men over women.
Unfortunately, sexism isn’t the only discriminatory bias AI has shown. An Asian student found out the hard way while renewing his passport on an AI-powered website:
AI bias in writing
AI bias and misinformation are similar to having a cult in which leadership and parents teach one view of religion and little math or science to their children. Without expert opinions and research, those kids will spread their views until a better, more trusted source proves them wrong.
Speaking of cults and fads, there are many excellent sources for AI information on Medium and elsewhere. There’s also an abundance of writers who have discovered variations of ChatGPT tools and are offering opinions provided by AI, and written by AI.
Increasingly, major publications — including Medium — are rejecting 100% AI-generated text and facts without sources, and Aman will as well.
Recycled content
When recycling plastic and paper first took off there were unintended toxic consequences. Recycling still has problems as the detritus needs to be separated from the fully recyclable clean products.
AI ‘knowledge’ is similar to what we recycle. We may need to do a lot of sorting, sourcing, and holding our noses to avoid contamination of facts.
Several of the major news and information-sharing platforms have opted out of allowing AI technology companies to use data from their sites. These include the New York Times and Medium.
AI-generated images and writing
I used an AI-generated image above to make a point. If you click on it and select “Search source using Google,” you’ll find other AI-generated images that are similar, but not the original inspiration for the style or content.
It’s quick, easy, and convenient to use AI for text and images, but the person who may have dedicated their time, resources, and talents will not be credited or compensated.
If I’d used an image from the artist Wyland — the living, breathing, non-compensated artist who inspired the image — I’d have credited him and he might have picked up more fans. I knew him in Laguna Beach, California, and Lahaina, Hawaii. He deserves recognition.
Wyland often created very large paintings and murals to grace the walls and sides of buildings in Laguna Beach, CA. Now they appear in several countries, sometimes enveloping entire buildings.
The unfamiliar observer might note many differences in style, but if you looked at Wyland’s body of work, you’d be convinced.
If you write a brilliant book, essay, or speech cited by many, you deserve recognition as well. Cheap imitations are rarely as good as the original.
Benefits of AI
There are many benefits of AI and reasons to be excited about future development. Anticipating and putting guardrails in place to ensure safety and unauthorized non-invasive use will help make the adoption of AI tools more successful in the future.
I usually write without an outline, then create a summary and ask ChatGPT or Bard a series of questions. In summary, “Here’s my intent and the likely audience. What did I miss? ”
Grammarly’s edit tool has improved my writing. (I recoil from red and blue underlines as if they’re the Priest in the movie The Exorcist — “The power of Grammarly compels thee!”)
When my AI hangover wears off I’ll be publishing tech-related work here and may do some guest-editing if needed.
I’m turning over the reins of Ethics & AI to Aman because I know him to be smart, supportive, and ethical. I trust others will discover the same.
Please welcome Aman Dasgupta and support him in Ethics & AI
Thank you, Patricia, for trusting me with the publication, and for all you’ve taught me.
Ethics and AI will remain a space to discuss the ethical concerns around AI technologies and ask questions about their impact on our future. With the breakneck (and reckless) pace of AI innovation, dialogue about “Ethics and AI” is the need of the hour.
The publication will continue rejecting 100% AI-generated content in favor of well-researched, experience-driven pieces about AI ethics. Please review the guidelines before submitting, and I look forward to reading all your wonderful submissions!
If you would like to write for Ethics & AI leave a comment with your profile name in the comments.
Thanks!
