New Frontiers in AI: The Next Generation of Enterprise Text Editors

I am definitely the last person to drink the kool-aid or jump on a new fashionable technology fad. People who know me would have often heard me say, “the people who know a piece of technology deeply are the ones who know not to use it.” So trust me when I say that there is a new frontier of AI opening up - it has always started off with my internal sense dismissing it as a fad until I’ve been convinced otherwise.?

In this quick write-up, I just wanted to share my thoughts on historical perspective which may be useful to readers anyway. Then at least show you one space in the industry that I believe will be disrupted.

Historical Perspective.?

Just to help with being useful, let me recap this world of advances the way I see it personally and I hope this perspective is useful for some.?

1. We had neural nets that came out way back in academic settings and they had tried to play chess, do reasonably well on speech recognition and then stopped being of serious interest for a long time.
2. We then had deep learning and out of that image and video processing. Everything started off with machines automatically identifying cats on YouTube videos. This was the beginning of deep neural networks that created waves of advances, mostly with academic or fun use-cases without serious industry applications.?
3. We had Apple come out with SIRI and at that time it was a bit underwhelming although the founder & architect of Apple SIRI is a close friend and advisor of mine and SIRI improved dramatically over time. Google and Microsoft followed as well.?
4. Amazon Alexa, Google home made serious improvements in both speech recognition and more importantly Q&A. IBM Watson beat the top jeopardy players with ease. Now, I use Google home on a daily basis.?
5. AlphaGo defeated the world’s best Go players using deep neural nets and I definitely recommend watching the AlphaGo Documentary. As a master level chess player, I can tell you that it is no easy feat at all.?
6. AlphaZero comes out with no pre-training. This has not gotten anywhere close to enough coverage but if you ask me, this is a far more serious technical and practical advance (and even scary) than AlphaGo which got a lot of press.?
7. LSTM based deep neural network models come out (prior to some of the above advances) that can start to do language comprehension in written text much better than what we could achieve before.?
8. Transformer-Encoder models come out that are simpler, faster to process and at least as accurate as LSTM models to do text comprehension and text generation. BERT and GPT-2, open-source packages come out as a result. OpenAI as the name suggests comes out as the open-source rival to Google, Amazon and Microsoft to ensure that everyone has open access to text analysis.?
9. OpenAI paradoxically comes out with GPT-3 that is not open-source but a serious advance over GPT-2 in both text comprehension, summarization and text generation. Other alternatives come out such as GPT-J, WuDao from China and now it is essentially a serious technological race. This is also one of those areas where you can throw money at the problem and make serious advances (as opposed to say breaking RSA which short of building Quantum computers is not clear what you could do). I am sure that nation states across the world, and definitely the US (NSA & CIA) have dedicated teams, products and deep budgets in-house to do exactly this.?
10. Where are the industrial applications? Google Home, SIRI, Alexa are great but niche. Google GMail auto-complete is seriously good, and just recently was made available on Google docs as well as Google slides. What else? I think this is about to change.?

When I was watching this space and dreaming up products, outside Google autocomplete and Grammarly, there wasn’t anything in the ecosystem mid 2020 - startups or otherwise. Now, there is conversion.ai, copy.ai, compose.ai (that I invested in) and more. I do not think people understand the change that can be brought. Let me give you a simple example.?

Next Generation of Text Editors.

Grammar and auto-complete are a given. But let's go further. If your text editor is used for corporate / enterprise use cases, should it not be a platform? Why wouldn’t your text editor talk to Outreach, HubSpot, SalesLoft, Marketo, Seismic, HighSpot and a whole host of other sales and marketing tools? Why wouldn’t your text editor talk to your ATS systems to pull and push job posts? Needless to say it feels like your text editor should certainly talk to your CMS systems, WordPress etc to write webpage copy. While we are at that, what about SEM copy and ads? Should your text analysis be done automatically by AI to detect inclusive language, brand consistency and messaging consistency? Should your text editor tell you what you should write, when you should write it and even help you write it? What about storing text in projects such as what Figma does rather than files and directories? It's a pity that if I write a 100 page document on a text editor today, there is absolutely no way for me to find all the countries I mentioned in it, the times I used font size less than 8pt font, the places where I used italics instead of bold. All of these have implications on accessibility & inclusive language. What about inconsistencies in those styles when I do have them across a 100 page document? Shouldn’t these be standard??

Hopefully I can give you a glimpse of what the future holds in terms of text creation, storage, and analysis and how AI can be deeply embedded not just inside your next generation text editor but also deeply integrated across the tools that use those pieces of text to communicate. I think there will be a host of startups that will disrupt this space in a variety of ways. It will be a lot of fun if we can help build one.?

p.s: this text was written and analyzed by a text editor in skunkworks :)