- Responsible by ClearOPS
- Posts
- Responsible
Hello, and welcome to Responsible, by ClearOPS. We started this newsletter because we believe in ResponsibleAI and responsible business practices.
Here’s what we have for you today
Have you wondered if Anthropic is really more responsible than OpenAI’s?
How to build your company’s AI policy.
Anthropic was created by ex-OpenAI employees who wanted a more safe, accurate and secure company. At first glance, I like it, but let’s dig a little deeper. The first question that I had was whether they treat my data differently for their consumer models (Claude vs ChatGPT).
Yes and no.
If you give Claude a thumbs up or thumbs down to a specific output, Anthropic can use that for any purpose, including re-training their models. In my opinion, Anthropic takes the “opt-in” approach to using your data for training. Conversely, OpenAI takes the “opt-out” approach. Personally, I do think opt-in is more privacy forward and ethically responsible.
But I was a little concerned that this disclosure was hidden in their online terms. So, for research purposes, I tried giving an output a thumbs up and Anthropic does indeed pop up an additional disclosure.
Thumbs up Anthropic!
P.S. Congrats to Anthropic on poaching a member of the defunct OpenAI superalignment team. More popcorn please!
In 2004, I was working at Gunderson Dettmer as an associate. My husband and I decided to take a vacation over Thanksgiving to Anguilla because, well, places in the Caribbean were relatively cheap over that Thanksgiving break and work was slow. It was awesome! We talk about that vacation all the time because it was just so lovely. The water is this beautiful crystal blue with a hint of green and the beaches are soft. The place we stayed, the CuisinArt (no longer exists) had cooking classes for me and dive fishing for my husband. But Anguilla was not as popular, nor did the island have as much money, as its popular neighbor, St. Martin. I believe it was because it did not have a functioning airport.
Apparently, Anguilla’s country code - .ai - means it controls the .ai domain name. The popularity of GenAI is making Anguilla some serious money, eclipsing its tourism revenue in 2023 (see the chart).
This is hilarious! How’s that for AI making the world a better place? Now, who wants to go to Anguilla?
I was speaking to a teacher this past weekend about how AI is affecting her job, specifically the rising popularity of ChatGPT by students. She told me that the school has set forth rules for students’ use. Specifically, that they can use ChatGPT as a tutor, but not the output.
To give this an example, if you write a paper and then ask ChatGPT to analyze the paper for passive tense and point out to you where you used the passive tense, this is acceptable use. If you tell ChatGPT to re-write the paper eliminating all the passive tense, this is not an acceptable use.
Brilliant.
As companies grapple with their own “do’s” and “don’ts” of using ChatGPT, Claude or Gemini, I think this proves that you must come up with custom policies that apply to your own business and industry. Want to figure it out together? I am interested in a few of these collaborations.
Speaking of collaborations, a reader suggested that we build a list of ResponsibleAI companies. What do you think?
Reply