Why am I so patient with Bing?
Charlie Dougherty
Executive Director of the Norwegian Peace Council Leadership and Growth | Antler 2021 | LSE ′09 | "The Best Papa" - Eva, 5 and Sunniva, 1
Why are we so kind and patient with AI assistants?on the internet?
I have noticed something really funny: If google only gives me ads or cheap content marketing on the first page of a search, I just close the tab.
If Forbes fills half the screen with ads, I close the tab.
If Linkedin takes too long to load, I move on.
However, if Bing or chatgpt give me some poor results or answer, I practically apologize for asking the wrong question. I was probably 12 years old the last time I asked something in a browser for something politely. Yet now at 38 I am taking up the habit again.
An example: I wanted to create an abstract image using bing.
Here is my prompt:
An abstract image in 3D using blue and green that represents the feeling of being in control
Here was the result:
I then tried to adapt my prompt, and asked to remove the human figure and make it abstract. DALL-E disagreed with me, and apparently it couldn't find a human figure in those images. So it returned in practice the same results as after my first prompt.
I kept trying and trying. I must have tried to improve the results at least 5 times, and this included an appeal to a friend who is good at this. to help. He told me to give the computer a really strong negative reaction to see what happened (nothing happened).
In the end I gave up. I couldn't understand what DALL-E needed, and DALL-E couldn't understand me.
Not only that, DALL-E took forever. Bing seems to display an answer just a little faster than I can type.
And every time I asked Bing to make an image it would:
Why, was I being so patient? Not only did I not get what I wanted, BING was astoundingly slow (in an internet context) to reply. Whats the deal?
We are never this patient with any other tech that I can think of. Developers are notorious in reducing load time for a website by milliseconds to not lose eyeballs, but I don't mind when Bing types out an answer just a little faster than I might.
My guesses are:
People have researched cute AI!
Xinyang Liv, Yue Liu, Jingjing Luo, Yuqing Liu, and Chunxiao Li report in the Annals of Tourism Research (87:2021) that "cuteness" has a positive effect on a customer or user's tolerance of service failure. They use cuteness defined as "... the ability of a vulnerable living entity to trigger the human innate desire to nurture (Wang, Mukhopadhyay, & Patrick, 2017)".
This leads to a sense of tenderness in the user and a reduction in performance expectancy. I dont find Bing and ChatGPT particularly cute. But is my patience a sign of nuture?
Are they cute, though, when they acknowledge their shortcomings and apologize for any mistakes? Is asking for feedback from the user cute?
Apparently I am a big sucker for manners.
领英推荐
However, the researchers did recognize that time pressure did limit the effects of cuteness. Like anything else, cuteness can only take an AI so far.
In this case, I ran out of patience. I also ran out of hope. I Just did not see how I could come up with a useful prompt.
So I just went to unsplash and found a free image in 30 seconds, downloaded it, and put it into LinkedIn instead.
It was really much easier than dealing with an AI helper, and the result was arguably better for my purposes.
I cut my losses with my prompting efforts and I realized that DALL-E was not going to give me 20X improvement over unsplash.
Maybe I am just a bad prompter. Maybe my vision is so dull stock photos are more appropriate to my purposes.
Regardless of how cute DALL-E is, it still does not do well with emotions.
One time I asked it to produce something "cheery" and there was a definitive motive of "cherries".
However, when I tried to recreate this "cheery" prompt for the sake of this article, this is is what happened:
Prompt:
can you make me an abstract image in green and blue that is cheery and optimistic
Result:
If you know what words here deserved to be censored, let me know. This annoyed me, and the reply was certainly not cute. "See how you can improve your prompt" puts the burden on me.
Of course I am reading too much into it if I feel that DALL-E was attacking my character, but I can't help find that when I know it wasn't my prompt, but it says it was my prompt, I do not desire to argue with a computer.
I look forward with both anticipation and dread to the day DALL-E knows my desires better than I do, but they day has not come, and it is not cute enough to keep me entranced like a siren on the rocks.
Every technology has to find a balance between potential and usability. Bing is still finding this balance, and I applaud them for putting out a product that is not polished.
Disregarding the general question of putting LLMs too quickly out into the world, I do think that the competition to provide services is letting us get access to these earlier stage services sooner than we otherwise would have.
I think that that is very fun. I also think that exposing Bing so early will make Bing better.
LLMs have so much potential, it would be a shame if they got lost behind overengineered user experiences that stifle creativity and dont leave any space for us users to help take the technology in unexpected directions.
The coolest results are the ones that we can't plan for.
Regardless, we can only hope DALL-E is still cute even when it is running the world.
*If you have read this far, leave me a comment or reaction and connect if we aren't yet connected! I write because I enjoy it and it connects me to others. Let's connect!