The most important thing AI has done so far is letting me click “Thanks” in Teams instead of having to type it out
AMD's newest generation of desktop processors do not have specialized AI cores, unlike their latest generation mobile chips. The reasoning is that the regular cores can be used to do AI stuff if needed. Hopefully this signals the end of unnecessarily making it a point to add AI "features" that actually don't add value to consumers!
I've been extremely happy with the free version of Claude 3.5 Sonnet. It's been able to do pretty much everything I've asked, including providing pretty good and accurate responses to complex tax law questions. I'm running draft emails and draft memos (redacted of course) through it more and more frequently to have it make them more concise or reworded for simplicity. Usage must have been high earlier today because it relegated me to 3.0, which was noticeably worse than 3.5. ChatGPT is getting blown away by Claude 3.5 IMO. I'll copy and paste the exact same prompt, and Claude provides "college-level responses" and ChatGPT is more akin to junior high, filled with obvious errors and inaccuracies. The tone/style of Claude is noticeably better, too.
I've been using them a decent bit locally, either through HuggingFace's APIs or things like Ollama to load the models locally and run it through cmd prompts.
Need to turn this into the tech / AI thread and discussion regarding how it is making our lives worse, not better Spoiler
This is possibly the dumbest argument in existence. Oh no, my dinner was ruined and totally worthless bc I didn't have the experience of saying "Yes, I'm done" then waiting for the check, slipping in a CC, then waiting again for them to bring it back so I could give a nod and a thank you! Pizza place near me allows me to pay via a QR code. It was different, but it really didn't ruin my meal bc I didn't get to slip a CC into a book, then wait 5 minutes to sign it. We should go back to cash bc of how personal the experience is!
Yes, outrage seeker who gets trigger by rando negative reviews....now mad about payment methods at one restaurant. We should bad resy too. The soulless quest of searching for available reservations has destroyed dinning. We must go back to endlessly calling restaurants to find one with an opening
i just started a graduate program and one person had a plaud, which is an automated dictation device that records lectures and converts it to notes https://www.plaud.ai/?srsltid=AfmBOorYtwfjk4cFRcATwNSKLvQt6e7WQ7dijf3QZ7Sg41iulAd0BIuP this person shared the notes in our team's slack and i'm blown away
THis is just a standard feature of Teams/Zoom/Meet at this point. The Zoom and Google Meet versions do an incredible job of summarizing meetings, so I'm assuming the Teams version does as well. I'm trialing something similar in our telephony suite to summarize phone calls so that when our call center agents go to document a call, they're just copying/pasting after a quick edit. Fairly useful so far. That's integrated with other full-on Co-Pilot features that suggest the next best action (change an address, update a bank account, that type of stuff) based on the ongoing conversation. Pushes knowledge content to the agents screen. Shit like that.
Is anyone paranoid about surveillance? It makes me wary that management could read anything that gets said in a meeting.
Every single keystroke on your machine is tracked. Everything you say in meetings is tracked. Even your mouse movements are tracked.
Also, monitoring a Teams call is one thing. But I've been at an in-person meeting in a conference room where the screen is on and dialogue is captured in captions on the screen. That seems like it could be used to listen/record a conversation that up until a few years ago you were pretty damn sure was off the record.
We just use it for some meetings and it only gets sent out to the meeting invitees. I'm sure the Google admins can probably get to the summaries if they really wanted to.
definitely depends on the organization I’ve tried to fire plenty of shitty employees for essentially not working and this data was not readily available or if it was the accuracy was questionable
I should have said “Assume that…” My org is extremely security forward and default policies on most machines include logging of pretty granular stuff.