1/ This is the state of #AI voice assistants at the dawn of 2020:

“Alexa, is it raining?”
“Yes it’s raining right now. Would you like to hear the forecast for the weekend in London?”
“No, actually, I’d like to hear the forecast for right now”

So great. I’m impatiently waiting for @RobinhoodAppUK to launch > @RobinhoodApp lets you invest as little as 1 cent in any stock https://t.co/vDCeULLsBS

The most important part of the Universal LEGO Sorting Machine is how @justasquid has used #AI. In this video, he explains the problem with using synthetic data for supervised learning and how to overcome it with domain randomization: https://t.co/LhTxsPz8jw

Finally somebody made it. Great job @JustASquid > Watch this machine made out of Lego sort other Lego using #AI – https://t.co/fiTwr5imR7

I never understood why @buffer doesn’t offer the ultimate tweetsplitting capability. I needed it countless times. https://t.co/q0g200WHv8

Chirr App is, by far, the best tool to split tweets I have tried (and I have tried a lot of them). Great job ⁦@kossnocorp⁩ https://t.co/wNpf2fdQ8x

@algorithmia 6/ E. Innovation hubs (AI-focused or not) only work as long as they serve the interests of all BUs. The moment they don’t, to innovate, each BUs is forced to deploy ML on their own, which doesn’t scale.

@algorithmia 5/ D. At least 25% of data scientist time is spent on deployment efforts < You cannot complain about this after you want DevOps/MLOps/etcOps. That task time doesn't go away simply because you changed ownership.

@algorithmia 4/ C. 18% of companies surveyed (745) take 3+ months to deploy an ML model in production. It sounds like the average time to deploy infrastructure before the advent of self-provisioning portals in cloud computing 15y ago…

@algorithmia 3/ B. 55% of companies surveyed (745) have not deployed an ML model in production yet. While there’s overwhelming choice for #AI enabling tooling, clearly there’s very little to help understanding #AI applicability to business.