I confused Google's most advanced AI – but don't laugh because programming is hard

We Keep you Connected

I confused Google's most advanced AI – but don't laugh because programming is hard

Most Popular
Care to have some fun? Fun at Google’s expense? And when I say fun, I mean fun in, well, a morbid fascination, what-were-they-thinking, I’m-glad-I’m-not-the-product-manager kind of way?
Our story begins with a Gemini Advanced release notes update published by Google on Tuesday. It talks about a new inline editor for the programming language Python:
What we’re going to focus on is the word “now” in the phrase, “you can now edit and run Python code snippets directly.” This seemed cool, so earlier today I set out to test it out in order to report back to you about this cool new feature.
Also: What is Google’s Gemini AI tool? Everything you need to know
Only…not so much.
Just a quick recap. Google’s version of ChatGPT (its generative AI) was called Bard and is now called Gemini Advanced. What Gemini is to ChatGPT, Gemini Advanced is to ChatGPT Plus, including the $20/month fee. I went ahead and signed up for Gemini Advanced, adding it to my collection of monthly AI-related bills.
Also: You can try Google’s new ‘AI Premium Plan’ for free. Here’s how
Gemini Advanced adds a bunch of new features. Note the highlighted line below:
Yes, it says, “Our most capable AI model.” Keep that in mind, along with the word “now” from earlier in this story. Here’s what the Gemini Advanced web interface looks like:
I went ahead and pasted in a simple Python script. This one simply creates and outputs the string “hello world” in upper case and sentence case.
Gemini Advanced first deconstructed my code and gave me a nice little explanation
It was then kind enough to give me the output of the script. Points to Gemini Advanced. The output was correct.
But ask yourself this: Do you see any way to edit that code? Any way to run the code as one might in a traditional code interpreter? No? Neither do I.
Obviously, I’m missing something. 
Now, let me be clear. I’m about to mock. I specifically asked my editor if I had permission to mock, because this is Google and we like Google. I was granted a full release to let loose and mock.
Also: Want to work in AI? How to pivot your career in 5 steps
I’ll try to be gentle.
Looking back at the release notes, there was this phrase, “Exclusive to Gemini Advanced, you can now edit and run Python code snippets directly in Gemini’s user interface.” So, my first question was whether or not I was actually using Gemini’s user interface. Here’s what I was told.
Notice, next to the arrow, a recommendation to go to https://www.gemini.com. No problem. Of course, Google owns the domain Gemini.com, right? No, they don’t.
Gemini.com is a crypto exchange. When I asked Gemini Advanced (which, don’t forget, I had to agree to pay an extra $20/month for) how to use Gemini’s Python interface, which was promoted in the release notes as “now” available, Gemini pointed me to a crypto site.
That, folks, is what we in the AI business call a hallucination. Oopsie. Yeah, and it gets better.
Also: How I tricked ChatGPT into telling me lies
When I pointed out to Gemini Advanced that it sent me to a crypto site, it apologized for the misdirection. Then it told me it couldn’t find the interface. And then it told me that what I was looking for might be internal to Google and might not be publicly accessible.
Gemini Advanced then ended that prompt response by asking me where I had originally read about Gemini’s ability to process Python scripts. Uh, Google’s own release notes. Duh.
By this point, I was feeling a Full-On Grade A Mockportunity forming in the cosmos. I fed Gemini Advanced the URL to its own release notes, and got this:
I want to point out some fun Gemini phrases (in italics below):
So, where does that leave us? Instead of an article showing off how you could dynamically edit Python inside of Gemini Advanced, you got this little romp through the verdant fields of “what else can possibly go wrong?”
Also: How Google’s AI helped me fix a Gmail technical problem
The feature in question shouldn’t be all that difficult to implement. Inline editors and interpreters have been found in coding environments for decades. Linking one into the Google Advanced interface should just not be that hard, even if there’s a need to pass some text between the interpreter and the AI.
To be fair, product and feature releases are challenging, especially for a company the size of Google. A tremendous amount of coordination needs to happen. Many teams and constituencies are involved. Everything has to go off with clockwork precision.
Also: How to take the mystery out of managing multiple projects successfully
As you can see here, this does not always happen — even if the company is Google. I’m sure the inline edit and test feature for Python will show up eventually. When it does, we’ll test it out. In fact, the company has announced that Gemini 1.5 (Gemini Advanced currently uses Gemini 1.0) is now in early testing. When that’s available, we’ll test it, too.
In the meantime, I’m going to hoist a mug of gloriously hot coffee to the Gemini Advanced marketing and engineering teams. “Here’s to you! It’ll go better next time.”
What about you? Have you tried Bard, Gemini, or Gemini Advanced? Have you signed up for either the added cost ChatGPT Plus or the Gemini Advanced subscription? What has your experience been? Are you looking forward to inline Python editing? Let us know in the comments below.
You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter on Substack, and follow me on Twitter at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, and on YouTube at YouTube.com/DavidGewirtzTV.

source

GET THE LATEST UPDATES, OFFERS, INFORMATION & MORE