Does AI make strong tech companies stronger? | Hacker News Reader

Does AI make strong tech companies stronger?

ben-evans.com

68 pointsmkbkn22 comments

So basically yes despite this author’s claim it will decentralize. It will make market leaders in any field stronger / less threatened as they have the relevant proprietary data.

This is why I vehemently disagree with anyone that claims creating a startup is easier today than in the past.

Not only are the low hanging ideas saturated (ie website monitoring, brand monitoring, fitness apps), but the bigger ideas require a lot of data, and the big tech companies have the clear advantage.

I don't think it is any easier or harder. The difficulty remains the same.

Today, the web stack is much more accessible, and accessible to a much larger group of people. Indeed, a non-trivial number of its parts are relatively automated (e.g. Weebly). In 1995, it was only accessible to a much smaller group of people. Mobile app development will soon reach the same point. And yet, there are plenty of new technologies and ecosystems - to learn and creatively generate something from - that are inaccessible to people in the same way the web was in 1995.

Experiment until you find something people want that has a promising business model and tech model design. If you get traction, don't take investment from just anyone (I'm reminded of Amazon's VC arm working with Amazon to copy a portfolio companies' product, which then buried that startup). Etc. Etc. There are plenty of other early-stage strategic heuristics you can utilize to your advantage.

Perceived weaknesses or disadvantages can be the other-side of an actual strength or advantage.

------

I want to emphasize this point more: often times, the lack of resources (whether financial, intellectual, etc.) is the setting of a creative breakthrough that generates the successful business model and tech model.

I think creating a small/lifestyle startup is easier than ever since distribution/globalization is much easier but creating the next big thing is much more difficult.

That's the best phrasing I've read that reconciles these two beliefs (easier to build + saturation). Nicely and simply put.

I think this is pretty scary when it comes to growing market concentration and inequality.

Basically, the most successful companies are the ones that can learn the fastest and apply that learning the fastest. Companies that already have lots of customers and data will be able to learn much faster (i.e. "get to statistical significance much faster") than those that don't.

Isn't it pretty uncontroversial that AI and automation will only worse inequality?

We can only hope that the inevitable revolutions to come will not be crushed by the all-seeing AI surveillance and AI-powered robot armies of our wealthy overlords before they even get a chance to become a threat to them.

Supreme Leader Xi of China is certainly hoping that this will be the outcome of the government's aggressive investment in AI technologies.

How could a broad generalization about the future, couched in terms whose definitions are controversial, be uncontroversial?

Agreed. It isn't just the tech industry, but banking, media and all industries. The market leaders in every industry with data, resources and scale stand to benefit the most. Which will incentivize mid or small sized businesses to merge or be bought out by large corporations. So rather than decentralization, AI will add to the momentum for greater centralization.

And looking beyond the private sector, the largest governments and nations with data, people and scale will stand to benefit most as well.

I think Ben Evans is a lot more optimistic than I am about about how feasible to train good models from data. By the time you have considered all the caveats, the number of candidates for applying ML greatly decreases.

I don't disagree that the number of applications of ML/AI is not as universal as the hype would lead one to believe at present.

However, don't underestimate that AI/ML could be a secular growth over twenty+ years. From that perspective, marginal (but constant) growth could result from widespread adoption of these methods (with tweaks) to new markets.

A VC should necessarily take a relatively long perspective,so your two viewpoints are not necessarily irreconcilable.

Lots of good, accessible points here, but I think it's easy to overestimate the moat around [business] data. [Google] data includes pretty much every aspect of human activity, since they keep their platform in your pocket, in smart home devices, and somewhere in the path of so many online intents. Chinese surveillance and shared platform data is a similar asset - both are everything-adjacent for a massive population.

And that's the big guys. A couple of years ago, this article might have said that MasterCard and Visa have all the spending data... But then Paribus proved that a scrappy startup with a free service could get tens of millions of people to share all their online receipts in record time, and give Capital One a great way of catching up through acquisition. That's not an equivalent dataset, but it's good enough for a lot of applications.

I definitely think proprietary data exists, and that companies will benefit be exploiting it with ML. But they should be very careful about assuming their data will uniquely cover an industry, or even a wide swathe of applications, for long. And they might not have to simply leverage ML, but actually reorganize their business around what can remain unique about their data (like they do about every other asset).

And for VCs evaluating new data-oriented startups, I wonder if they will need new thinking about the time horizon on which investment pays off. Once an application for a new data asset proves valuable, it may turn out to be much more replicable than expected.

>>> but actually reorganize their business around what can remain unique about their data (like they do about every other asset).

this.

I am advocating the idea of a "programmable company" that is the end point of automation - where once you find product market fit the rest is automated - perhaps a better phrase is market / data fit

It's not AI itself but data and hardware tech they don't share (freely or commercially).

That’s what they’re saying, just like sql, ML will be accessible to everyone and it’s having the data to make use of it that will differentiates companies.

I enjoyed the article but I see two overlooked aspects that significantly weaken the author's argument:

1. Statisticians are substitutes for data. You don't necessarily need new / more data if you have a statistician.

2. Data often contains a lot of redundant information. Big data may simply be duplicative.

Could you elaborate more on the first point? I don't really get it

Humans can often learn new things from many fewer data points than current ML algorithms. Sometimes a single data point suffices.

Human statisticians can apply a variety of mathematical tools to fit different situations. ML systems tend to be more like one-trick ponies.

I’m not sure I could agree with that statistician can “learning new things” with smaller data points. Statisticans might come and see the pattern with “better” prior than ML models, that allow them to come up with better conclusion. However, given the same dataset, there is a maximum to thee information that can be extracted from the dataset. Ideally, Any human(think of human brain as a pattern recognizer) or appropirate statistical method would come up with the same information from the datast, given no prior

Amazingly there is no upper limit on information that can be extracted. :-) Also, by applying increasingly sophisticated techniques, statisticians can extract increasing amounts of information from the same data. Don't have data? Just add a statistician.

I’m not sure if you’re trolling or not, cus there is definitely an upper limit to how information can be encoded in data.

It will be interesting to see if the strong tech companies are willing to sell their data via API or just safe guard it for their own use.

Reply on news.ycombinator.com