The good news is, we have managed to make compute a commodity (although not a cheap one), and to greatly reduce the need for specialists via frameworks like TensorFlow and Keras. Removing bottlenecks one at a time
Show this thread
The two bottlenecks in the deployment of ML: - Availability of large quantities of labeled data - Availability of ML engineers Large companies have a decisive advantage on both fronts. This is why we're unlikely to see a new ML startup skyrocketing to a position of dominance
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Not sure if it's bottleneck in this specific framing or just another element of today's ML reality, but I think distribution (to existing customers, users) is overlooked and a larger reason you may not see ML startup skyrocket.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
How about a startup inside Alphabet?
-
just check out
@monkeylearn, amigo ;) - Show replies
New conversation -
-
-
@fchollet Sad but true!Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@VentureShire this is the common point of view in ML industry :) I think that just saying “it’s hard” is too easy. It is probably correct for most cases, but several others there might be a better answer.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
IMO data is in abundance, the need to label it properly for the specific intelligence is again an ML engineer skill set. So if tht analogy is correct then it boils down to one thing : shortage of “Good” ML engineer.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Strongly agree. At http://scaleapi.com , we're working to make high quality data labeling an infrastructure service that every company has access to. The 2nd might be solved with some AutoML-like thing offered by cloud providers.
-
Id just go with Mechanical Turk
End of conversation
New conversation -
-
-
In EU I would add as bottlenecks: - Availability privacy lawyers (GDPR) - Availability infrastructure (big corps not always keen on the cloud)
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.