AA 017: When will the deep learning bubble burst?

TL;DR: There is no deep learning bubble

Welcome to the fifteenth podcast of the Ask Alex show! Today the question is on “When will the deep learning bubble burst?”

In this podcast, we talk AI, the deep learning bubble and all the BS going on:

• The scale of the opportunity from AI
• The challenges we face
• Calling BS on startups using the AI term and the marketing BS told by Watson

Thanks for Listening!

To share your thoughts:

• Leave a note in the comment section below.
• Ask a question over on Ask Alex.
• Share this show on Twitter, Facebook, or Pinterest.

To help out the show:

• Leave an honest review on iTunes. Your ratings and reviews really help and I read each one.
• Subscribe on iTunes.

Transcript on a deep learning bubble if you prefer to read

To reply to this prima faces is to accept the premise. Is there actually a DL bubble? And furthermore is the bubble pertaining to DL or the broader bastardised notion of ‘AI’ which is better characterised (for now) as ‘advanced analytics.’ DL is a subset.

I believe the application of AI, as defined, has yet to even touch the surface. The commercial impact on a broader service industry that will support real impacts from enterprise to SMEs is immense.

At a very micro level of application, I see a huge industry popping up in script writing once chatbots actually work. The professional services in implementing, service and maintenance etc will be extensive. I have some research papers that have theorised the value of the service industry of AI. It’s big money.

The applications of a lot of ‘ai’ has been severely limited to affluent organisations (Huggable  octopus, or not?). Your Google, Baidu, Facebook, Microsoft (the most) and IBM have been hoovering up talent. I think there is basically only one uni in North America, ie Canada, which has been able to retain senior Professors (there was an a16z podcast around Jan 2017).

Sure some startups have talent but it’s limited in scope. The temptation of acquisition is omnipresent. So long as that persists and the potential addressable market size and cost savings and revenue maximisation ops persist there will not be a reduction in investment, which one would presume is the yard stick for the bubble?

I’ve done a fair amount of research into the space, particularly in enterprise, and the implications are insane. There is no bubble. Just a lot of BS.

Where there is a bubble is in the vapid misuse of AI-related terminology. Ergh, gross. The pitches that say they apply ML to… and are a ticketing company by any other name is silly. It’s the SoLoMo of the 2010s. Sonar etc flamed out as battery life etc was an issue. Snapchat buys Zenly – Your friends, live. for 250m a decade later as they solved geolocation related geolocation… or whatever was actually publicly announced.

Watson is causing issues to the industry with the total overhyping of their tech (presumably as they need to move the needle on a titan(ic) to make investment appreciable to share price). Great, they have an amazing marketing machine, but it’s all hot air. The hot air bubble is Watson. They are persona non grata in Singapore having failed so terribly with the gov and major banks on more than one occasion. They’ve been told not to apply to any more RFPs….

The pragmatic opportunities are highly verticalised. This notion of wide-AI is going to let everyone down that thinks that is what DL/AI is and is coming anytime soon. Also, ‘DL’ is one facet. For many high-value solutions requires many disciples to pollinate, including NLP, ontology, reasoning and the many forms of perception (scheduling and possible robotics, depending on application). All these other disciplines of ‘AI’ are not talked about at all.

To get nerdy, I theorise we need to apply DL to automate development in ontology and NLP building to push the needle, but anyway 😉

The bubble is multi-fold in pretenders who are hopping on the band wagon to cover inadequacies in their value prop, per se. We simply haven’t seen any real value outside the major corps, so I see a massive depth to come. I think Hyperscience in NYC is a cool example to come.