From his sources:
KI Expenses rise to 2.5Trillion Dollar by 2026
Gartner predicts up to 40% of Enterprise Applications will include integrated task-specific Agents by 2026.
The animation he made about the Restaurant Call Agent is quite good. The problem is that we are are here building stupid systems with AI by design default. I do not mean what it does, but how. My point is, that the system he explains will every single time when being asked to make a reservation call start a full reasoning process instead of take an existing successful one and exchange known variables.
This causes in Enterprise Applications extreme inefficiencies and might cause easy fixable issues to remain being solved by a Standing Work Around, not to mention hard to predict Butterfly Effects.
A smart AI Agent needs to create a database of it successful solutions to by that speed up its responses, guarantee success and be predictable beside deliver insights of its use. In Server uptime, to pull a related topic, post comma figures are important onto mission critical levels being measured and actively monitored:
The true "uptime" of Apache is often measured by the community rather than a corporate guarantee. In the late 90s and early 2000s, it was common to see Apache servers with "uptime" stats like this:
Metric AchievementTypical High-End Uptime
(99.999%) (approx. 5 minutes of downtime per year)
Common Configuration
(99.9%) (industry standard for most businesses)
Read that carefully again and now wonder what 5% hallucinations will do in your company in a Standing Work Around your employees got used to out of convenience.
I got a free of charge provided coffee while my Computer booted up. Every single morning. 8 years in a large U.S. corporation booking not one single second break time.
How much does Amazon loos in 5 minutes a year? Enough to get a second server being a perfect duplicate and you want your AI Agent forget its thoughts...each single time.
#cyberpunkcoltoure
#opensourcecommunity