5 Google IO Updates for Business Owners | New Gemini Updates in Google Workspace

[@gerzio1]

Excellent video!

[@itGenius_]

Thank you very much!

[@IAmSam_TheMan]

Google needs new leadership and it needs it bad and it needs it ASAP. i’m a huge Google fan but I am so tired of being outdone by Apple, Microsoft and now ChatGPT. Can some please change the leadership at Google? look how Microsoft turned around with its new leadership. Great Video though! 😇

[@itGenius_]

yeah... seems like new energy is needed around much of their product ecosystem or perhaps just the challenges of a huge global corporation it's now become..

[@DonMikimax]

This thing should be included in google workspace, not as an addon. For tokens, i agree, WTF are tokens, nobody knows!

[@DonMikimax]

Google services are so complicated like their help page. It feels like whoever is making help pages is a complete different company, completely disconnected from the products. I can never find what I'm looking for and it always feels like the help page is wasting my time.

[@AFTC-LTD]

Calling ML, AI was the biggest mistake ever. Marketing hype. Having tried many times to train gemini, claude and gpt on documents in various formats none of them can follow prompt instructions or document information supplied correctly. Best you can get is its early response being correct, but after a few paragraphs into a response and it starts to do its own thing. So much for larger context/token windows and memory capabilities. Then you end up asking it to correct its response which is no good for people who dont know the LLM training documentation (subject matter) really well. So be careful using it on long email conversations. Simple or individual specific tasks only for a good few years...

[@itGenius_]

It’s definitely getting better and better, though I agree I’ve also some really frustrating experiences with very simple tasks. Gemini is the worst offender for me

[@AFTC-LTD]

@@itGenius_ I had a great example of this from google's new Notebook LM, I fed it a single paragraph explaining what the document was about and about 10 to 15 short points of information on the subject, then clicked generate pod cast. It started off quite impressive, very good quality voices and interaction between the two presenters, very human like, but the further into the podcast it started talking about subjects and mentioning facts that weren't in the doc I fed it A lot of WTF moments🤣 Claude has basically a huge hidden prompt behind each question you ask it, GPT does the same but in different ways, their new release o1 just prompt engineers itself to give a better response. If you have used stable diffusion you will be familiar with prompt crafting to get the right result, this is the same thing. GPT o1 while improved for doing some self prompt engineering (the fake thinking it has going on), it's no smarter than GPT 4o, and asking it the same question (not a fact or singular answer response question/task) will often give you a different response/answer each time.