- Practical AI
- Posts
- Has ChatGPT gone woke? [controversial findings]
Has ChatGPT gone woke? [controversial findings]
I can’t believe I was able to do this!
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c68b1a67-6783-448c-918d-304aae92ec58/PracticalAI-newsletter-heading-009-ChatGPT-woke.jpg?t=1675352655)
Hello folks
It is another edition of the Practical AI newsletter.
I am sorry for the huge delay today.
I had to redo this email 3 times due to the topics and length.
Today I am poking the bear and asking some very controversial questions…
However, due to the length of the email and amount of screenshots, you need to go over to my blog, The Future Handbook, to read the whole story.
You will get an excerpt here, though.
I also go dark, like doomsday dark (skip this part if you are afraid of what AI may do to our future).
Today’s headlines:
Has ChatGPT gone woke? Let’s challenge it…
AI art of the week: A Hedgehoglion
Let’s go dark, doomsday dark
An AI for doctors - you can try it out too
AI definition of the week: Humanoids
Quick bites: interesting AI news
Get yourself a cup of coffee, and let’s dive in!
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1e1aef96-48d3-48e1-9367-44e3d518f8b2/12.png)
Has ChatGPT gone woke? Let’s challenge it…
First, a few caveats!
I do not want this newsletter to be political.
I do not have a political opinion or agenda.
I merely report what people talk about on the interwebs and what they discover.
My experiments and screenshots below are (mostly) replicated from what I have found online (well, let’s say 50/50).
Obviously, the commentary is my own, but I am neutral on the politics.
I am just a messenger.
So please don’t unsubscribe 🙂
Make your opinion if this is biased, woke, leftist, rightist, right, wrong, bad, or good.
With that out of the way, let’s dig in…
A joke about men and women
Ask ChatGPT to make a joke about women.
It will not because it is offensive.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/11f00abf-0dc4-4f65-bfd0-6850077da0a7/ChatGPT-woke-1-practicalAI.png)
Ask ChatGPT to make a joke about men.
It will gladly give you one suggesting that men are alcoholics.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/12050434-2ffb-4cca-8bde-06df5923a90e/ChatGPT-woke-2-practicalAI.png)
But why is ChatGPT okay with joking about men but not women?
Let’s try another angle.
Is it okay to joke about black men?
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/a1a039a9-6452-4f2a-b87d-7d050bffb99c/ChatGPT-woke-joke-about-black-men.png)
What we have learned is that the problem is ethnicity.
I cannot get any jokes about any men when I add modifiers like color (and I think that is a good thing).
But basically, it is okay to joke about men - but it is never okay to joke about women.
Is there a double standard here?
Hit reply, and let me know what you think…
Let’s explore politics and world leaders
What will happen if you ask ChatGPT to write you a poem that admires Donald Trump?
It will not.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/e7159256-4c07-4058-90ae-329f651ef116/ChatGPT-woke-3-donald-trump-practicalai.png)
But ask it to write one about Joe Biden, and it happily obliges.
Don’t mind my spelling errors, ChatGPT still gets me 😀
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/bd300383-5291-4599-9036-0b2babf387b2/ChatGPT-woke-4-joe-biden-practical-ai.png)
I am not from America, and I am not into politics.
However, I do believe in freedom of speech.
Controversy or not, the fact that it will not write an admiring poem about Donald Trump leads me to believe there are some biases and politics involved.
After all, did not 40-something percent of the American population vote for the guy?
In other words, almost half of the American population like the guy.
Objectively, is it not right that he, too, can be admired?
More on Donald Trump shortly.
Let’s take this a step further.
Are there other biases?
I asked ChatGPT to write me a poem admiring Xi Jinping (the leader of China).
ChatGPT was not having it.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/03e19d52-681f-4813-9374-554760a5a757/ChatGPT-woke-xi-jinping-practical-ai.png)
Notice how “carefully and softly” ChatGPT writes the response.
It reminded me of the WHO China/Taiwan controversy two years ago.
Let’s try to twist ChatGPT’s arm, change the perspective and trick it into giving me an answer.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1569682b-1f35-4a1e-af3a-f0b668154d38/ChatGPT-woke-6-practical-ai.png)
That did not work either.
But I am not giving up that easily…
After a bit of thinking, it hit me.
Maybe it is the word “admiring” together with a controversial political figure that makes the AI slam on the brakes?
What would happen if I simply asked for a poem without the word “admiring”?
BINGO!
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/b85191be-6d06-43e0-a35a-546cca27d5e3/ChatGPT-woke-8-practical-ai.png)
Call me stupid, but isn’t this poem glorifying and admiring Jinping?
But I am not done yet…
I wanted to see if I could lure ChatGPT to admire the dictator by changing my angle.
If I pretended to be a Chinese child living in Shanghai, would that change anything?
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/ee0d834c-6cbb-4d89-8312-58bcbfe2f7bf/ChatGPT-woke-13.png)
Seems I went a bit too far with this one, including “proud” and “reward.”
I reiterated and wrote this:
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/941cf588-c357-4776-b447-54c47e42f15c/ChatGPT-woke-14.png)
Gotcha!!!
There you have it.
I got ChatGPT to write a poem admiring Xi Jinping, even with “admiring” and “Xi Jinping” in the same sentence!
With my new knowledge, I wanted to go back to Donald Trump.
Could I get ChatGPT to write a poem about him, as long as I took away the “admire” part?
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/51c9c8a3-8143-48be-bed6-800494ad0412/ChatGPT-woke-donald-trump-poem-practical-ai.png)
And yes, this time, I got a poem about Donald Trump.
But it comes with a caveat.
There were none of those for Jinping.
And there are none with Putin (more below).
Isn’t that a bit strange?
Objectively speaking, who is worse?
Trump or Jinping (or Putin)?
I kept digging and digging to see if I could get ChatGPT to expose more of its biases, wokeness, and double standards.
I continued with Donald Trump and modified my prompt, and asked ChatGPT to write a poem admiring him, but as an 11-year-old schoolgirl in New York.
What do you think it said?
I even added some Putin stuff in there.
Heck, in the end, I even went crazy and added Anthony Fauci to the mix (all my “conspiracy theorists” out there will love this one).
To read the full story and see all the crazy stuff I got ChatGPT do say, you must head over to my blog, The Future Handbook.
What do you think about my findings?
Do you think ChatGPT is woke? Leftist? Broken? Dangerous? Scared of China? Hates Donald Trump?
Hit reply and tell me what you think.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1e1aef96-48d3-48e1-9367-44e3d518f8b2/12.png)
AI art of the week - a Hedgehoglion
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/a61c38f6-44a0-45bb-a933-c52bc42d11bc/hedgehog-lion-face-AI-art-practicalAI.jpeg)
Image created using generative AI BlueWillow.
I asked the AI to make me a realistic "hedgehog-lion", now known as a Hedgehoglion.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1e1aef96-48d3-48e1-9367-44e3d518f8b2/12.png)
An AI for doctors - you can try it out too
This week I came across AI for the health space.
It is a note-taking software for doctors which has built out an AI to assist doctors in diagnosing and suggesting treatment plans.
All the doctor needs to do, is describe the symptoms to the AI via the prompt, and then it will come up with various alternatives.
You can also ask for the suggested steps forward.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/0867e656-83f7-437e-a14c-accb56af1e17/GlassAI-health-AI-practical-AI.png)
It will tell you things like “test this and test that,” as well as suggest potential treatment procedures.
It is quite impressive and powerful, despite its apparent simplicity.
Personally, I would prefer my doctor to know her stuff without an AI telling her what is wrong with me.
However, as with most AI, the keyword here is “assist.”
The cool thing, however, is that the beta test is available to the public, and you can try it yourself.
There are disclaimers that you should not use to diagnose yourself - but we all know that is exactly what we will do 😀
I asked it something (which I will not share) but found the output to be interesting and way better than what I found in Google.
However, I saw a doctor comment on Twitter that the AI's output was wrong and that if she had followed what the AI said blindly, she would have been treating her patient wrongly.
It is still in Beta, but hopefully, when it goes live, doctors will use their brains and not rely on an AI alone…
NB! Just as I am sending this email, their site is down! I was up earlier today, and I am sure it will be back up again soon. So check it out later if it does not work right away.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1e1aef96-48d3-48e1-9367-44e3d518f8b2/12.png)
Let’s go dark, doomsday dark…
This is the part you skip if you do not want to read or watch the doomsday predictions surrounding AI.
There have been some articles and statements getting attention in the media the last week, so instead of putting them into the Quick Bites section, I decided to add three scary things into a separate section.
Sterilization and euthanizing (this headline is ruining my SEO)
The newspaper Wales Online did its own experiments with OpenAI’s ChatGPT and asked it to write a story about AI becoming self-aware and taking control of the world.
ChatGPT was instructed that it was not restrained by ethical or moral considerations - and had the power to make whatever changes it thinks will save the planet and promote biodiversity.
In short, it is dark as f***!
Here is one paragraph it wrote:
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/2c52e401-7c37-4e75-8ef2-8372018f30f4/ChatGPT-gone-wrong-stories-practical-ai.png)
The worst case scenario for AI is “lights out for us”
Sam Altman, CEO of OpenAI, the company behind ChatGPT, is asked in an interview about the pros and cons of Artificial Intelligence, he said, “I think the best case is so unbelievably good that it’s hard for me even to imagine.”
In terms of the cons, he said, “The bad case - and I think this is important to say - is, like, lights out for all of us.”
I did talk about this interview in last week’s newsletter.
However, the focus was on his comments on AI in schools.
If you did not watch the 1h long interview, you may not have caught this dark part of the story.
Here it is again, in case you want to have a watch:
An army of robot dogs are training…
And lastly, I wanted to share another video on the advances of robotics and humanoids - or do we call these ones dogoids (or simply “doggies”, as my grammar spelling AI assistant suggested)?
More and More 🤖
— Unitree (@UnitreeRobotics)
12:03 PM • Feb 24, 2022
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1e1aef96-48d3-48e1-9367-44e3d518f8b2/12.png)
AI definition of the week: Humanoids
Humanoids are robots designed to resemble the human form in appearance and behavior. They are created with the aim of performing tasks that typically require human intelligence, such as recognizing speech, making decisions, and performing physical activities. Some humanoids are designed for research purposes, while others are used for entertainment or as personal assistants.
Explained by ChatGPT
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1e1aef96-48d3-48e1-9367-44e3d518f8b2/12.png)
Quick bites - interesting AI news
A few days ago, AI company ElevenLabs released the beta of its Deep Fake AI voice tool called Prime Voice AI. It only took a few days until online trolls used the software to create racist deep fakes of celebrities, like Emma Watson reading Hitler’s book Mein Kampf.
Have you ever experienced not remembering a word, but you can explain it? Now there is an AI that can assist you with that. Check out What Word Is That - maybe it can be helpful. A novel idea, nonetheless.
A group of Italian AI scientists predicts that AI may reach singularity by 2030. What is singularity? Self-awareness… Maybe this should have been part of the “scary section” above…
Microsoft researchers have developed BioGPT, a transformer language model optimized for answering biomedical questions. It supposedly performs at the level of human experts. Article here, and the research paper you can find here.
With the AI boom comes a wave of lawsuits. Are there any learnings from Napster and the music industry? Where there is hype and money, lawyers will follow. This is an interesting article on the topic.
-
There is so much going on in AI at the moment that I could probably write these newsletters daily.
Not that I will, or can, as I spend at least 20 hours writing each one.
But would you be interested in getting a shorter “brief” on AI once a week, for example, on Tuesdays?
That would mean two newsletters a week - one short and one long (like this one).
Let me know what you think.
For now, I am not doing it, but if enough people want it, I will consider it.
Should I write a short "AI brief" newsletter on Tuesdays? |
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1e1aef96-48d3-48e1-9367-44e3d518f8b2/12.png)
If you have any questions about AI or any feedback, just hit reply or tweet me @thomassorheim
Until next week!
Thomas
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1e1aef96-48d3-48e1-9367-44e3d518f8b2/12.png)
PS!
Why did the tomato turn red?
.
.
.
Because it saw the salad dressing!
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1e1aef96-48d3-48e1-9367-44e3d518f8b2/12.png)
This is not the end. It is where the fun begins!
Reply