How well do I know my industry?
This seemingly simple question is actually quite difficult to answer. As an entrepreneur and CEO, one would hope that the answer is “very well” – after all, it’s common wisdom that a big part of success for startups is the founding team, and particularly how well they understand the market they are going after.
But how on earth would you measure something like that?
In most cases, you don’t need a precise measure. Investors aren’t looking to see whether you have 100 pips of knowledge – they’re looking to see if you have enough, for whatever fuzzy, subjective bar they happen to have set. Different investors might disagree about exactly how knowledgeable a founder might be, but even those imperfect assessments would be enough for their purposes.
The thing is, though, what if I don’t just want to signal that I know what I’m talking about to investors? What if I actually care about being right that I know my industry well – and especially about ensuring that I’m gaining useful knowledge over time rather than losing track of what’s going on?
Well, another way to think about “knowledge” is “that which constrains your expectations.” When you know nothing about a subject – say, chemistry – you anticipate that anything could happen. (Most humans know something about chemistry, though – we might not know whether to expect two liquids would bubble, emanate a gas, or do something else, but we probably know they won’t turn into a clown and start singing. This is an important point for those analyzing cognitive processes philosophically, but for my purposes, I’m just going to designate a universal baseline of knowledge about how the world works – for instance, clowns don’t just magically pop into existence – and discuss knowledge beyond that universal baseline.)
So, the more I know about a subject, the better I should be able to constrain my expectations about what will happen in that area. In other words, I should be able to make narrower, more confident, and more accurate predictions than the next guy or gal – all three of which criteria can often be assessed objectively. This isn’t a fast process – you can’t evaluate the quality of the predictions until the expected time comes to pass – but I believe that the process of making predictions and testing your performance is crucial in order to ensure you’re continuing to grow. (One might argue that this is a reason why investors, especially those with many investments, might have a better sense of how a market is evolving than e.g. a loyal reader of the Wall Street Journal would – even if they are consuming the same information, the investors are making predictions all the time in the form of their investments, so they’re forced to more directly confront when their beliefs were wrong, and update them accordingly.)
So, keeping this logic in mind, I’ve spent the last few years making predictions at the beginning of the year, calibrating my confidence levels, and evaluating how I did at year-end. The past few years, these predictions were mostly focused on either my own – personal and professional – life; Modulate’s specific success; and important topics which come up in day-to-day discussions like politics. For a variety of reasons, I’m not going to share these predictions publicly here. But this year, I’ve decided to add a variety of predictions about gaming, synthetic media, and the overall tech industry, and I thought it would be valuable to share them. So without further ado, let’s dive in – looking forward to hearing your thoughts, and if you’re willing, your own predictions as well!
(One last note – I’m avoiding publishing any specific predictions that directly relate to Modulate or our voice skins product, and instead trying to focus more broadly on the industry at large. This is mostly to avoid confidential info, but also because most of my predictions about Modulate are a bit more on-the-surface – if I wasn’t confident voice skins would take off, I wouldn’t have cofounded Modulate in the first place!)
Prediction: Other than voice skins, some 50M+ MAU games release significant new features which are tailored to encourage players to participate in the game’s voice chat.
Probability: 70%
Context: Voice chat is becoming more important as games become more and more social. While voice skins are certainly one feature, I expect games to come up with a variety of innovative ways to encourage people to spend more time with their in-game community.
Prediction: Free-to-play game revenues increase by 5% or more over 2019’s totals
Probability: 60%
Context: As games become communities moreso than simple experiences, it becomes more important to increase the player base – just as social networks like Facebook only are valuable if enough of your friends are also there. Free-to-play is one of the easiest ways to increase your player base – but the test of whether or not those players stick around is whether they eventually start spending money in-game. (This is lower confidence due to uncertainty about the impact of the broader economy, rather than uncertainty about the trend overall.)
Prediction: Total XR revenues will still be less than 25% of PC game revenues alone (as measured by SuperData’s year-end report.)
Probability: 80%
Context: XR revenues (which even today were mostly driven by e.g. Pokemon Go, more than experiences which fundamentally rely upon VR or AR) are certainly creeping upwards, but it seems that we as a community have yet to build the toolset to really make the most of this new technology’s capabilities; and much less to convince consumers to seriously invest in exploring it. I think there’s a lot of exciting stuff to come here, but we’re probably still a couple years from proper takeoff.
Prediction: There is no technology in the world capable of translating the language you are speaking in under 100ms during a live call.
Probability: 90%
Context: There are fundamental limitations to how much a machine learning system can understand about the context of what you’re saying if it has to work on the fly instead of waiting for you to finish your sentence – and super high-level aspects like translation are one of the hardest problems out there. I won’t say it’s impossible (and it could even be Modulate that cracks it…) but I think most people underestimate how hard this is compared to other problems neural nets are starting to grasp.
Prediction: A single text-to-speech platform takes off, finding adoption with greater than 5% of game developers for at least some NPC voices in game.
Probability: 50%
Context: I’m pretty skeptical of this – voices are really nuanced, and I think TTS will need to push further into giving the customer control of those nuances before it can really take off. That said, there’s a lot of interest in this kind of thing, and gaming has enough centralized tools with e.g. Unity, Unreal, WWise, etc that I could see one of them offering a solution that starts getting picked up more earnestly.
Prediction: No social media tool with a userbase at the size of Instagram, Snap, Facebook, etc today releases an automatic-deepfake-alert feature which alerts those browsing the content (rather than just flagging uploads for a moderator to examine).
Probability: 80%
Context: Unfortunately, synthetic media can be created in a variety of ways, and detectors are becoming less reliable and less generalizable as the tech keeps pushing forwards. I do think it will be possible to have something like this down the road – if we’re able to e.g. set a good standard for watermarking your media so it’s identifiable, etc – but I’d be pretty surprised if it turned out these platforms could develop a reliable, fully-automated detector without that kind of starting point.
Prediction: At least two organizations outside Google will announce Stadia competitors. Whether or not this happens, I further predict that none of them, including Stadia, reach player bases which are meaningfully competitive with the big console names.
Probability: 70% for the first sentence, 60% for the second
Context: To be honest, I still don’t quite get why Google chose to invest so much in Stadia – which makes me think I’m missing some important insight around the business model or the world they’re aiming for. But I suspect that, if they felt this was worth trying, others will follow. That said, either way, I’m still not seeing the consumer world latch onto this technology – and there are a number of extremely hard technical problems which are yet to be solved. If Stadia does succeed in a big way, I have trouble imagining it coming from anything more than Google’s sheer force of will…though that’s certainly not impossible!
Prediction: Multiple released games will appear as viable candidates for an eventual online “omniverse” – somewhat akin to Ready Player One’s OASIS, though significantly narrower in scope and technical viability to start.
Probability: 70%
Context: There’s a ton of interest in this kind of comprehensive online space right now – ranging from games like Fortnite transitioning to a kind of social hangout, to VR games focusing on coexistence and explorations, to many games in development which received some sizeable checks this past year. I don’t expect us to be close to the real thing yet, but I’ll be surprised if we don’t see some really direct and material progress.
Prediction: Investors who currently label themselves “AI-focused” will split into a couple different camps, drawing lines between “big data predictions/analysis” vs “automating tasks” vs “building creative tools”, among other things. We’ll see this trend in the evolving vocabulary used by funds and companies moreso than through anyone explicitly rebranding themselves.
Probability: 50%
Context: This one is kind of tricky. As AI research continues to produce more and more powerful tools, the field of “AI and data science” has grown and branched in many significant ways, making it ever harder for any one person or even firm to understand the nuances of all the different branches. Combining this with the overall trajectory (though we’re still early on it!) of certain kinds of AI/ML becoming more utility-esque (though “the new electricity” might be a bit much) while other orgs use AI/ML for more central or novel means, we’re going to see questions about what an AI company is that mirror “is WeWork a tech company” today, forcing people to clarify their focus. I feel pretty good about this overall claim…but I’m not certain how far we’ll get in this coming year alone.
Prediction: ???
Probability: ???
Context: I’m sure I’ll think of additional predictions, so I’m leaving myself room to add more – feel free to suggest things you’d be interested in my thoughts about in the comments! Of course, I’ll make sure to denote when, if at all, I add these predictions, to avoid “cheating” from predicting things after seeing us get halfway there.
That’s all! Before I close this, though, a few notes for those of you who are sticklers for the rules.
1. I’ve framed all these predictions such that my predictions is 50% or greater that the proposition is true. This is so that I can most efficiently check my calibration at the end of the year – saying I think something is 20% likely vs 80% likely is an equally strong claim, in the same way that saying “there’s absolutely no chance that happens” is an overconfident assertion, in contrast to the underconfident “I don’t know if there’s a chance that would happen, so I can’t make a useful prediction one way or another.”
2. Not all of these predictions are trivial to evaluate. I opted to leave them loose rather than binding myself to a particular way of evaluating them – for instance, someone claiming “I think the economy will improve” could surely find a measurement by which the economy appears to grow even if most pundits agree the economy is hurting. At the end of the year, I’ll publicly discuss my evaluation of these claims – both to solicit feedback about whether I’m evaluating something unfairly, and to force myself to engage with them as honestly as possible.
An Entrepreneur's Journey 