AI Chatbot Explorations ??????

An exciting gamechanging lift-off. Realism sinks in. The future is still bright.

tldr:

#ai has a long way to go before it's mainstream: chatbots can't even count correctly (at times) or excel at #sentimentanalysis. Still, the future is very bright & exciting ?? ?? ?? . There are use cases (see 1 example below) that could blow our minds ?? and help non-techies excel in a digital world. ??

Next LinkedIn post, I will show you the #pythonprogramming needed to use #apis to access OpenAI's #llm so that you will say, "if Hank can do then certainly I can do it!" #aiprogramming is a possibilty for all of us.

Like many engineers at heart, I am coming up to speed on chatbotting #ai. I am doing so by using APIs to access #llms like OpenAI's #chatgpt v3.5. On the other side, I am exploring prompt engineering: comparing OpenAI's #chatgpt v3.5 to Anthropic's #claudeai v3.

Here's some of the fun developments

• #chatgpt3 can interpet a "play on words" or puns very well, but jokes centered on human insecurities is a bit hard for it. See pic and chatgpt v3 session link for more info... (btw, #claudeai has the same limitation with understanding human sentiment (at times).

• #chatgpt3 has a hard time with counting (yes, simple math) (see pic below). This is a known limitation: some times we liken this gap in #aiknowledge to Albert Einstein, who understood the universe, but couldn't tie his own shoes ?? (not sure if that story is true ????♂? ). Interestingly, though, #claudeai does not have this counting limitation or problem. It counts, properly. 10 words -- no problem!

• In the end... Regardless of how far along #aimodeling has to go, the future is (hopefully) bright ? ?? It is a game-changer.

Simple Use Case Example: The Help Desk of the Future

Can you imagine #aivideo technology, where you prompt,

"my computer is broke."

• A video of your exact computer (Mac, PC, Linux) appears and you're asked to point where the issue is.
• Then it troubleshoots for you... showing you what to do via an easy to follow video of your exact computer ... down to the correct color and shape of your computer!
• If a resolution cannot be achieved then the bot automatically creates a ticket for tier3 support, bypassing tier1/2.

This would/could speed resolution and reduce frustration. Another example is allergies in food. You take a pic and the AI reverse engineers the pic and tells you if you might have an allergic reaction to the food. Anything could be possible is the point!!

Ethicists like Timnit Gebru and others raise excellent points. Matter of fact, their expertise is this space is beyond reproach. I am amazed at how well they understand the challenges of #predictiveanalytics (more of the same inequalities at scale ???? ). So, moving slow is moving very fast!

What's Next

• Here is a link to my entire session including the chatbot misunderstanding the joke and counting words (if you want to examine the details)
• My next post will be of the Python code to use #chatgpt #apis to prioritized unstructured datasets and files before the primary #llm. The use case I will explain, too.
• BTW, this has been done a 1,000 times, but now an old-time coder is showing you

"if I can do it then anybody can do it."

• You can transition to #aicoding.