Top 50 Collaboration Industry Thought Leader, Evangelist and now Analyst - Engaged with UC, AV, Multimedia, Video, and AI... A general technology influencer, storyteller and force-multiplier.
This is my point about #AI and meeting summaries. Below is a promo that I recorded this morning for my Monday podcast/webcast. Watch the video, and turn on LinkedIn’s automatic captioning. According to those captions, I’m going to a basketball event and I’m asking everybody whether it will be covered on CNN. Even though the captions are 90% accurate, those little mistakes completely change the gist of what I’m saying. It is those speech to text captions that feed your meeting summaries. So if you’re reading or sending out automated meeting summaries with these tiny little critical mistakes, you’re risking your entire business on something that is clearly inaccurate. #ItsNotReadyYet!!!!
Top 50 Collaboration Industry Thought Leader, Evangelist and now Analyst - Engaged with UC, AV, Multimedia, Video, and AI... A general technology influencer, storyteller and force-multiplier.
Join me on Monday's #UCWeeklyNews when I'll be in Las Vegas for the 2024 National Association Of Broadcasters conference. ~10am EDT here and basically everywhere - but you can always go directly to Dantoreport.com National Association of Broadcasters #NABShow #NAB2024
Unbelievable!
Top 50 UC Expert. AI Show co-host. Leader BCStrategies. Analyst/Consultant for orgs and vendors.
4moDavid Danto. Balance is a good thing. Many are overhyping AI but to declare, universally ,it is not ready, is to swing the pendulum to the other extreme. Individual test cases making interesting articles / posts but are not representative of the general state of AI transcription. Just like when you take a drug and rave or rant about it, this is not a clinical trial! I am finding it hard to cite specific studies comparing AI to human transcription accuracy because all the articles (studies?) I find are either sponsored by companies offering humans to transcribe (they say AI sucks) or by AI firms (they say humans are more expensive and less available). Here are a few facts: - there are different real-time and post-recroding transcription models; one values speed the other values accuracy - accurate transcription incurs more compute and thus sometimes you get what you pay for; i.e. free transcription services are quick and less accurate - results vary based on many variables: devices, gender (most models are still more responsive to deeper voices), accents - AI will likely rapidly improve; faster than any previous technology (given both interest/hype but more importantly corresponding investment) Let's keep the discussion going!