Brutal takedown of shift from "big data" to "AI" startup narratives

[ChpBstrd]

https://www.youtube.com/watch?v=pOuBCk8XMC8

TL;DR -

• From the early 20-teens until recently, startups sold themselves with a narrative of "big data", "algorithms", and "data science". The promise was that these tactics would lead to insights, disruption, and profits.
• Instead, many of these startups like Groupon or Wish went on lose billions of dollars over the next 5-10 years. Many are gone today.
• The video stitches together lots of clips of corporate representatives essentially repeating a script about big data and analytics. It is very unnerving from the point of view of an investor.
• After a decade of "big data" and "machine learning", margins and other metrics at regular companies remained stable. There was no effect.
• The current AI narrative is merely a rewriting of the old big data and algorithms narrative. It explains why the previous narrative didn't work by saying the data was too big for humans to comprehend, and we need machine helpers to generate the insights.

[MustacheAndaHalf]

I stop watching YouTube videos (like that one) as soon as I realize they are anonymous.  I don't see an indication of them being "big data" companies in their Wikipedia entries, but I'd still like to understand your definition for big data.
https://en.wikipedia.org/wiki/Wish_(company)
https://en.wikipedia.org/wiki/Groupon


If big data is combining the work of many computers, data centers might be required for big data.  The companies with the largest data center footprints are also the big tech companies, so I still need an expanded definition of big data to make a complete list.
1. Amazon  +27.6%/year for 10 years
2. Microsoft  +26.6%/year for 10 years
3. Google  +20.0%/year for 10 years
4. Meta   +22.1%/year for 10 years

[swashbucklinstache]

I'm ignorant except working at a small analytics company.

@MustacheAndaHalf, fair to say that except meta those are the companies both selling shovels and mining for gold? It is tough because cloud adoption has several very strong reasons of which big data is just a small part.

The optimist in me would say most of big data was about enabling companies like AirBnB, doordash, instacart, microlending etc. The pessimist in me says it might've moved the needle directly for like 20 large caps and maybe 2,000 micro/startups contracting in to help them. Then there are an unlimited number of VPs who just rotated from low code/no code to digital transformation beyond the core to big data and now to AI with the hopes that if revenue grows 15% during the quarter they have Their Initiative they can use to take the credit and get promoted to SVP before never thinking about it again.

[ChpBstrd]

The point is these companies sold themselves as investments on the theme of "big data" and "machine learning" a decade ago, and their efforts yielded no results.

The video does point out that the people who sold proverbial shovels, like Amazon, did much better. Cash from investors flowed through the unprofitable "big data" companies like Yelp or DoorDash and into the pockets of people selling web server space and storage space for all those data.

Also, I would consider all YT videos equally valuable whether the creator is anonymous or not. First, even if the video starts with "Hi, I'm [firstname] [lastname]..." you don't necessarily know if that is true. It's also hard to check the references and credentials of a talking head, unless they are an academic, politician, or otherwise famous, and even then there is the deepfake problem. Third, even if the talking head was crazy enough to use their real name and we could somehow trust that they are who they say they are, I would suggest a selection of such individuals would be skewed in a certain direction - maybe toward narcissism, maybe toward a moonshot goal of becoming a celebrity, maybe with a book or seminar to sell, maybe just reckless or naive about the risks of the internet, etc. and such systematic differences could affect their perspective about various subjects. E.g. maybe many of the non-anonymous Youtubers are advocating real estate investments because they have something for sale and want to build trust to sell that thing, and the anonymous Youtubers are the only ones offering skeptical insights.

I think good or bad ideas can come from many sources, particularly when they are simple connections of ideas like the ones discussed in this video. It's the idea that matters more than the source.

[MustacheAndaHalf]

Also, I would consider all YT videos equally valuable whether the creator is anonymous or not. First, even if the video starts with "Hi, I'm [firstname] [lastname]..." you don't necessarily know if that is true. It's also hard to check the references and credentials of a talking head, unless they are an academic, politician, or otherwise famous, and even then there is the deepfake problem. Third, even if the talking head was crazy enough to use their real name and we could somehow trust that they are who they say they are, I would suggest a selection of such individuals would be skewed in a certain direction - maybe toward narcissism, maybe toward a moonshot goal of becoming a celebrity, maybe with a book or seminar to sell, maybe just reckless or naive about the risks of the internet, etc. and such systematic differences could affect their perspective about various subjects. E.g. maybe many of the non-anonymous Youtubers are advocating real estate investments because they have something for sale and want to build trust to sell that thing, and the anonymous Youtubers are the only ones offering skeptical insights.

I think good or bad ideas can come from many sources, particularly when they are simple connections of ideas like the ones discussed in this video. It's the idea that matters more than the source.

If you can't look up someone's credentials, that is a problem.  You claim people lie about their names, or their credentials don't exist, or if they use their real names they have bad intentions.  How is that different from a conspiracy theory, that everyone with real names is lying or cheating?  I didn't see any evidence mentioned, and I don't plan to pursue your speculative beliefs about real names without evidence.

Outside YouTube, there are journalists who always use their real name, because they stand behind their work.  There are researchers who ethnically must use their real names.  When I find articles and research, I often find someone's background - I don't see the claimed problem you're having.  And by seeing their area of expertise, I find out if they know what they're talking about.

Some writers lack investment expertise.  I've seen one mix up crashes and recession signals in a way no expert would do, and I only caught their false assumption owing to my personal experience reading dozens of investment books.  In areas where I'm not an expert, I don't want to take on the false assumptions non-experts can make.

It's unfortunate most of your post keyed off my very first sentence, instead of trying to create a constructive definition of "big data".  If you think there are only 2 big data companies, Wish and Groupon, and both failed... I disagree.  I listed 4 other companies to start a discussion around big data, but instead I got a long list of excuses why real names can't be trusted and have bad intentions.  A more constructive conversation would try and find all big data companies and how they turned out.

[GuitarStv]

I stop watching YouTube videos (like that one) as soon as I realize they are anonymous.  I don't see an indication of them being "big data" companies in their Wikipedia entries, but I'd still like to understand your definition for big data.
https://en.wikipedia.org/wiki/Wish_(company)
https://en.wikipedia.org/wiki/Groupon


If big data is combining the work of many computers, data centers might be required for big data.  The companies with the largest data center footprints are also the big tech companies, so I still need an expanded definition of big data to make a complete list.
1. Amazon  +27.6%/year for 10 years
2. Microsoft  +26.6%/year for 10 years
3. Google  +20.0%/year for 10 years
4. Meta   +22.1%/year for 10 years

The data centers for Amazon, Microsoft, and Google are used largely to sell cloud hosting - not really for the companies themselves.  Does hosting servers for other companies count as 'big data'?

[swashbucklinstache]

How about this for big data company?
A company that, through big data technology, can consider every or significantly more members in it's very large population for analyses instead of a sample, with very little limit to characteristics/variables it can consider or ways it can consider them. Doing so gives them a profit-driving competitive edge in an existing field or creates a new field. This is likely because on-demand usage costs of cloud tech lowers the investment cost and risk or the tech to handle the scale exists at all for the first time.

This removes shovel sellers, industries with small data, and industries where small sample survey statistics are entirely good enough. But leaves it obvious that people could not realize those are the buckets they're in.

Computer vision products are one easy example of big data "creating" new industries, if we're allowed to include machine learning. Auto vacuums, lawn mowers, ring doorbell etc. Maybe predictive things, like how full airplanes are or how good we can predict weather?

A this time is different to consider: do more accurate predictions reduce downside risk even if they don't grow profits, in such a way that moves the natural P/E higher? My company doesn't make any more profit than we did a decade ago but we make 20 widgets for every 1 we did then.

[ChpBstrd]

It's unfortunate most of your post keyed off my very first sentence, instead of trying to create a constructive definition of "big data".  If you think there are only 2 big data companies, Wish and Groupon, and both failed... I disagree.  I listed 4 other companies to start a discussion around big data, but instead I got a long list of excuses why real names can't be trusted and have bad intentions.  A more constructive conversation would try and find all big data companies and how they turned out.

The video covers Uber, AirBnB, Coinbase, Snowflake, Doordash, Rivian, Lyft, Opendoor, Pintrest, Robinhood, Affirm, Snapchat, Groupon, Pandora, Grubhub, Yelp, Zynga, Wayfair, Chegg, Warby Parker, Sofi, Casper, Blue Apron, Allbirds, Stitch Fix, TrueCar, Lemonaide, Lending Club, Fitbit, Roku, Oscar, GoodRx, and Instacart, plus a lot of bigger companies jumping in on the narrative like Pepsico, WalMart, Goldman Sachs, Boeing, The Gap, JP Morgan, Experian, Statoil, Paypal... and the list goes on. The video notes that the big FAANG companies generally captured the profits from that era and its speculative investment boom.

In each case, the evidence presented are things company representatives said, media the companies produced, or their financial results. So it's less editorializing by an unqualified vlogger, and more "let me bury you with evidence produced by the companies themselves". That evidence is verifiable, if not confirmed by recent memory. The contribution coming from the author is limited to illustrating how the hype of the past looks a lot like the hype of the present.

I think that simple claim, in the context of evidence presented, does not require as many credentials or fame to be persuasive, as opposed to something like "I have an econometric model that predicts inflation" or "I think this particular component manufacturer is technologically years ahead of their competitors" or even "stock XYZ is about to go down/up". For those claims, some analysis is required and the claimant either needs to show their work or persuade us to trust their work. This video is a far simpler connection of ideas, and the work is shown by the gathering of evidence.

Anyway, I thought it was worth watching for the insights we can gain by watching hype being produced in a historical context, as a frame of reference for understanding today's hype. It's not an academic piece and never could be. Yet I don't blame you for trying to place a quality filter on the torrent of information facing us.

[swashbucklinstache]

But I mean ... Some of those companies are profitable and clearly wouldn't exist or wouldn't have disrupted their market without big data. Of course they're going to hawk the latest and greatest to acquire funding. Some of them intentionally aren't profitable. That said, you can clearly see some on the list having no real business investing in big data except for the VP syndrome. What would you expect a VP who pushed big data and saw no profit to do, step down?

As an investor that seems in line with every tech hype thing we've seen before and will see again. It's probably going to be shovel makers, FAANG, and a few winners in a sea of losers. Sometimes from within the house it's real easy to spot the latter group. I'm reminded of the NoSQL fervor leading to a presentation from a (then) industry player who was going to Change Everything by using Mongodb. In an industry whose defining characteristic was how structured their data is, was, and will be.

[Fru-Gal]

I am so happy to see the narrative turn against Sam Altman now. More and more people are seeing him as an evil figure rather than humanity’s last hope against what he said was the extinction-level threat of the AI (that he was building). It’s still unclear what all the reasons behind the OpenAI board kerfuffle were with firing Altman, but it’s starting to look like what they said was the reason — he’s not trustworthy, a liar — was the reason.

And perhaps there is no AGI looming.

[WayDownSouth]

https://www.youtube.com/watch?v=pOuBCk8XMC8

TL;DR -

• From the early 20-teens until recently, startups sold themselves with a narrative of "big data", "algorithms", and "data science". The promise was that these tactics would lead to insights, disruption, and profits.
• Instead, many of these startups like Groupon or Wish went on lose billions of dollars over the next 5-10 years. Many are gone today.
• The video stitches together lots of clips of corporate representatives essentially repeating a script about big data and analytics. It is very unnerving from the point of view of an investor.
• After a decade of "big data" and "machine learning", margins and other metrics at regular companies remained stable. There was no effect.
• The current AI narrative is merely a rewriting of the old big data and algorithms narrative. It explains why the previous narrative didn't work by saying the data was too big for humans to comprehend, and we need machine helpers to generate the insights.

Without even needing to watch the video (which I will out of sheer interest), your bullet points are dead on. AI will be a huge failure. The term AI being tossed around doesn't even fit the definition. Google search is and was "AI" since its inception - aggregation and organization of data, which is then applied in a strategic manner depending on what it's been programmed to do. There is literally no "Artificial Intelligence" about it. It's human made, human controlled, etc.

Even if you give AI the power to make decisions, it's still doing so based on A.) it's programming and constraints, and B.) the data it is allowed to absorb and the data that's available for it to absorb.

Example: AI will never be able to to something such as engineer on its own. It always requires deep data. It will never create anything that's new, never been said before, or that hasn't already been concluded/hypothesized by someone else. It's nothing more that the power to compute. Our brains, while they can compute, also are interacting physically and emotionally with a 3-dimensional world and having a genuine experience. Computing is a tool and always will be a tool. It will not have a life of its own nor will it solve the worlds problems - any of them. It will however, make for cheaper and faster labor in many industries, as well as provide comforts that make people more lazy and passive.

I'm pretty passionate about this subject and was researching AI deeply since like 2008 when it was more "future speak" than anything else, and the talk was all about its potential. Negative, positive, whatever...

Could a military, Terminator (the movie) type scenario could happen and machines could take over and destroy everything? Only if humans with evil intent and enough power wanted that to happen and use AI as a scapegoat. No computer or group of computers is systematically going to create it's own robot army suing high technology to determine the human race must be destroyed unless someone set those parameters, not only within the technology, but within the socioeconomic landscape and allowed it to continue doing it's own thing without simply pulling the plug. It's powered by electricity. It can always be disconnected.

To say "it just happened and we can't stop it" would be complete bullshit. AI is pure hype. Make the $ on it in the market while you can, by 2030 it will be understood that it's flat-lined and what we have is simply robotics with a futuristic label and appearance.

[Ishmael]

https://www.youtube.com/watch?v=pOuBCk8XMC8

TL;DR -

• From the early 20-teens until recently, startups sold themselves with a narrative of "big data", "algorithms", and "data science". The promise was that these tactics would lead to insights, disruption, and profits.
• Instead, many of these startups like Groupon or Wish went on lose billions of dollars over the next 5-10 years. Many are gone today.
• The video stitches together lots of clips of corporate representatives essentially repeating a script about big data and analytics. It is very unnerving from the point of view of an investor.
• After a decade of "big data" and "machine learning", margins and other metrics at regular companies remained stable. There was no effect.
• The current AI narrative is merely a rewriting of the old big data and algorithms narrative. It explains why the previous narrative didn't work by saying the data was too big for humans to comprehend, and we need machine helpers to generate the insights.

Without even needing to watch the video (which I will out of sheer interest), your bullet points are dead on. AI will be a huge failure. The term AI being tossed around doesn't even fit the definition. Google search is and was "AI" since its inception - aggregation and organization of data, which is then applied in a strategic manner depending on what it's been programmed to do. There is literally no "Artificial Intelligence" about it. It's human made, human controlled, etc.

Even if you give AI the power to make decisions, it's still doing so based on A.) it's programming and constraints, and B.) the data it is allowed to absorb and the data that's available for it to absorb.

Example: AI will never be able to to something such as engineer on its own. It always requires deep data. It will never create anything that's new, never been said before, or that hasn't already been concluded/hypothesized by someone else. It's nothing more that the power to compute. Our brains, while they can compute, also are interacting physically and emotionally with a 3-dimensional world and having a genuine experience. Computing is a tool and always will be a tool. It will not have a life of its own nor will it solve the worlds problems - any of them. It will however, make for cheaper and faster labor in many industries, as well as provide comforts that make people more lazy and passive.

I'm pretty passionate about this subject and was researching AI deeply since like 2008 when it was more "future speak" than anything else, and the talk was all about its potential. Negative, positive, whatever...

Could a military, Terminator (the movie) type scenario could happen and machines could take over and destroy everything? Only if humans with evil intent and enough power wanted that to happen and use AI as a scapegoat. No computer or group of computers is systematically going to create it's own robot army suing high technology to determine the human race must be destroyed unless someone set those parameters, not only within the technology, but within the socioeconomic landscape and allowed it to continue doing it's own thing without simply pulling the plug. It's powered by electricity. It can always be disconnected.

To say "it just happened and we can't stop it" would be complete bullshit. AI is pure hype. Make the $ on it in the market while you can, by 2030 it will be understood that it's flat-lined and what we have is simply robotics with a futuristic label and appearance.

I've also been watching 'AI', and am a professional Software Dev who has studied the theory behind 'AI' a bit. I think many of your arguments are valid, but are being a bit too dismissive of it.

AI is not the right term, but it does model the way the human brain works. We don't really know why it works, just that it does. That's different than how systems are built now, and how I assume Google works, which are executions of algorithms, ie prescribed approaches to solving a particular problem.

While it works like a human brain does, it doesn't have the same base initiative/drive/consciousness behind it, so it isn't really "intelligent" - it's only following a "goal" that we give it - kind of like when we tell a dog to fetch a stick we throw. It is also not close to being flawless, as no human brain is, or can possibly be.

On the flip side, it's able to digest and incorporate MUCH more data than a human brain can. While general tools like ChatGPT are impressive but mostly cute, applying the technology in a focused manner to specific things is likely to bear significant fruit in the near future, IMO. One example to illustrate would be in diagnosing of medical conditions. Would I trust an "AI" to correctly diagnose what ails me? Fuck no. Do I think it could be an amazing tool for a doctor to run diagnostic information/symptoms through and suggest possibilities to consider? Abso-fucking-lutely - I'm sure there are people who bounce around in the medical community for years until they happen to run into the right specialist that correctly identifies their obscure/non-standard illness. So I think combinations like that will result in important, tangible, positive outcomes. I think the way Apple is incorporating it into their OS/apps - i.e. helping to accomplish specific things - is a good example, too.

Also, keep in mind the growth of technology is exponential. While the past history might not look impressive, IT technology usually follows a long lead time of looking that way, but then reaches an inflection point and grows in capability much more rapidly.

So my "prediction" is that we'll start to see tangible outcomes from "AI" that augments human efforts in the short term (i.e. increased productivity), while we collectively refine inputs and results. Then, we'll start to layer/integrate it, to have "AI" layers that leverage more specialized "AI" subsystems in ever-increasing abstract domains. At some point, it will be considered irresponsible to not follow the guidance of these systems without justifying a darn good reason why not.

After that, someone will give a more generalized AI a goal function, and the world will turn into a sci-fi story (utopia? dystopia?). The points I've read and which I find interesting to contemplate are:

• When we start making generalized AI systems with its own goals, how can we ever be sure that we've adequately covered/protected against all possible negative outcomes?
• There will be a moment when artificial intelligence is roughly equal to human intelligence; however, assuming it will be growing exponentially, that will likely only be a brief moment. After that, it will quickly become more intelligent than us. There is a relatively small difference in intellectual capabilities between us and chimps, yet most of what we do must be totally inscrutable to a chimp, because it's impossible to comprehend a significantly greater intelligence than one possesses. Therefore, the actions of the AI will most likely be the same to us - totally inscrutable.

Of course I'm projecting pretty far out into the future (I think?) but it all raises some pretty interesting questions, IMO.

[roomtempmayo]

Is there a point where "founders" start to be held to adult standards of truthfulness and delivery?

The video argues that both venture capital and founders have an interest in maintaining short memories and readily forgiving past bad behavior.  I'm not sure I understand why that interest is mutual.

The only real parallel framework I have is to think about research grants in academia.  The principle investigator is something of an entrepreneur who is pitching ideas to funders.  Sometimes the grant works out as planned, sometimes not.  But what's nonnegotiable is that the application be honest, and that the grantee do what they said they were going to do.  Violating those two principles would mean being blacklisted on future applications at best, or losing your job and having to pay back money at worst. 

I don't understand at all why in the world of tech and venture capital it seems to be just fine to fabricate stories out of whole cloth and misuse funds.  People then get to just say "oopsie" and be forgiven?  It makes zero sense to me.

[ChpBstrd]

Is there a point where "founders" start to be held to adult standards of truthfulness and delivery?

The video argues that both venture capital and founders have an interest in maintaining short memories and readily forgiving past bad behavior.  I'm not sure I understand why that interest is mutual.

The only real parallel framework I have is to think about research grants in academia.  The principle investigator is something of an entrepreneur who is pitching ideas to funders.  Sometimes the grant works out as planned, sometimes not.  But what's nonnegotiable is that the application be honest, and that the grantee do what they said they were going to do.  Violating those two principles would mean being blacklisted on future applications at best, or losing your job and having to pay back money at worst. 

I don't understand at all why in the world of tech and venture capital it seems to be just fine to fabricate stories out of whole cloth and misuse funds.  People then get to just say "oopsie" and be forgiven?  It makes zero sense to me.

Blame Bill Gates and Steve Jobs. Both were salesmen of paradigm shifts in the way we use computers and the internet. Both sold their visions using buzzwords, trend names, and vague jargon. Money learned to chase such behavior, and a million copycats were born.

[MustacheAndaHalf]

Groupon had a "series D" funding round for $950M in Jan 2011, and then went public at the start of Nov 2011 for a market valuation of $17.8 billion.  Even the VC investors who invested the same year as the IPO made money.  And investors who got in much earlier made a much greater profit.  You can't look at post-IPO performance and claim pre-IPO investors did badly.

https://www.startupranking.com/startup/groupon/funding-rounds
https://techcrunch.com/2023/03/31/groupon-which-has-lost-99-4-of-its-value-since-its-ipo-names-a-new-ceo-based-in-czech-republic/