Is AI in the Midst of a Hype Cycle?

Since we are still on this side of the hype cycle I have to ask the question: is AI really different than previous hype cycles? I have only been investing for less than a decade but I can remember at least 3 distinct hype cycles: 3D printers, blockchain, and cloud. At the time all you had to do was mention one of those words and your stock skyrocketed. They were touted as changing the world. It seemed as if they would never drop in price and the continual gains would continue into eternity.

At some point people begin to realize that the item within the hype cycle has utility, but the utility is not as far reaching as expected/touted. This usually results in the price pendulum swinging too far in the other direction, creating more rational buying opportunities.

There are a lot of smart people here who have been through a lot of these hype cycles and know a lot about AI. You always feel like this one is different, because it is by definition. Is AI really different? Or will people realize it has great utility but won’t change the world?

I’m sure there will be opinions on both sides of this, and usually there are very good reasons why this is different, but then only in hindsight it becomes apparent what was missed in that discussion. My goal here is just to hear people out because I feel at the moment like Nvidia will never go down, which to me is a surefire sign we are in a hype cycle.

-Junomean2

19 Likes

The hype cycle is about stock price. There is no one AI stock, there is a huge variety of businesses that produce or consume AI.

The hype cycle is about human emotions, the fear of missing out [FOMO], for example. Many people just read headlines, or watch Jim Cramer so their input is about the technology darlings, such as Nvidia, Tesla, and Super Micro Computer. Some important beneficiary companies go under the RADAR.

For reference, I started out as a programer at IBM in 1960 and had the opportunity to see early iterations of AI such expert systems that all essentially failed to the point that AI was declared dead. All these early applications were based on heuristics, that is, programs, algorithms, created by humans and totally different from how intelligence works. They had to fail. My work as a programmer created a working model of the brain. The model consisted of two brains, the lizard brain that most living things have, and a rational or boolean brain that distinguishes humans from most other animals. I call the lizard brain the subconscious which, somehow manages to come up with solutions apparently without thinking, an absolute necessity in a fight or flight situation. The heuristics I mentioned above model the boolean brain which is only part of intelligence, an insufficient part.

How does the lizard brain work? In my mental model it is a pattern matching machine that stores vast amounts of data (humans, under the right circumstances, can recall almost all they have ever experienced). See or hear a lion and you don’t reason about it, it’s instant fight or flight. In my mental model the lizard brain matches the sight or sound to stored patterns and kicks out the relevant ones really fast. That’s exactly what neural networks do. Only recently have neural networks become practical even when they are much smaller that even animal lizard brains.

BTW, Andrej Karpathy, who used to be Tesla’s head of AI, has said that the capacity of AI scales directly in proportion to the size of the neural network it runs on. This is a very important data point for AI investors.

o o o o o o o o o o o o o o o o o o o o o o o o o

If scared of the hype cycle stay clear of technology darlings and search out companies less likely to be FOMO driven. A most important data point is that size matters. Big computers, fast computers, low energy consumption computers, and deep pockets. Also companies that have other means of support, not just AI, to rely on.

My current daring, besides Tesla, is ARM Holdings. They are a platform for developing microchips. ARM powers most smartphones, Nvidia and Tesla use ARM technology. ARM beat out Intel because their RISC architecture is very power efficient vital in mobile applications as well as in data centers.

I’ve been an ARM follower for over 20 years. Some of my ARM articles since 1999

https://www.google.com/search?hl=en&as_q=armhy&as_qdr=all&as_sitesearch=softwaretimes.com&as_occt=any&safe=images

ARM tends to be under the RADAR because it was acquired by SoftBank and only recently has it been reissued.

Denny Schlesinger

38 Likes

Trying to organize my thoughts for the umpteenth time in order to reply.

I think I called the boombust hype cycle on AI last year, and it has just grown even more over time. So, I am totally on the side of, this is tulips all over again. (No, I am not old enough to remember that, but I am old enough to remember fiber optics in the internet boom.) 3D printing might be the closest hype boom/bust to what I believe is happening now.

Yes, there is a place for products using AI. Will it end up “in EVERYONE’S house” and “EVERYONE will make whatever they can possibly dream of!!”? I doubt it. So where are we on 3d printing, well all the hype built out the ideas and the tech to where some companies can 3d print rocket nozzles on many of our popular launch systems. 3d printing is starting to gain traction on housing actually. But is there any “money” to be made from the tech right now? I don’t see any. You can only invest in the companies that might be “using” the tech. (I have a bit of RocketLab stock.)

For AI, I am looking to the companies that are providing the support of AI in the same way that I watch the EV space for companies that are not just car companies. I don’t want to own a car company, no matter how many models they have or how fancy the doors are. I want to own the company “changing” the car world. Tesla is in that camp, but Fisker and others never were. (I even kicked Rivian out of my port recently for acting too much like a car company.)

I have two stocks under the AI umbrella. Nvidia is a support/hardware part of AI in that they don’t have to put money into a bot or a car, but they can make anyone’s bot or car better. SoundHound is my other, pretty much direct AI play. I own that one for the very base reason of ‘playing the AI hype’ and I would be happy at a buy out. Their work seems to be super important in the world of interacting with AI. The thing that killed 3d printing was that you still had to learn all the convolutions of software, programming, design, interations, etc. No fully employed human was going to dedicate time to the second full time job of learning 3d, just do they could print out dice or some cool bolt and nut design. Same with AI…right now you have to be a Google, Microsoft, Nvidia, Alphabet to have the resources and the dedication to making AI do what you need. How many hours of work goes into getting AI to walk up and down stairs? Now take them home and try to get one to walk up and down the wonky stairs in YOUR house…nope.
Same with AI art, or with AI generated music, or with AI generated product/search suggestions. They won’t work for most and would need to have a lot of fine tuning to get them right. I have literally be telling FB that i do NOT want to get suggestions for (American) football fan pages. I have NEVER been a huge football fan…but will FB’s algorithm give up? NOPE. How about FB’s instagram trying to see what shorts I watch…"Oh you watched on clip with a dachshund? Well here are a MILLION suggestions and only suggestions for dachshund related stuff for a week.

Long bit of rambling, but yes I agree that the hype is huge on AI and the FOMO is getting out of hand. So I am staying relatively out of the AI hunt, but if a company I already own says they are using AI for something, then I am good. I am not actively hunting every new AI company or trying to ‘get in on the ground floor’ on things remotely related to AI.

12 Likes

I’m sorry but this is just bonkers. I can understand extending the hype analogy to some of the other players in the market right now. But NVIDIA? The NVIDIA that just grew revenues 265% YoY, 22% sequentially? And is already profitable? The NVIDIA that grew earnings 765% (not a typo) YoY and 33% sequentially? I’ll wait while you do a back of the napkin math on annualized rates. Forward P/E, btw, is in the mid 30s. High for a hardware company? Sure, we can definitely quibble with that. All hype? Not a chance

Show me a 3D, blockchain, or cloud stock that ever came close to sniffing those numbers (and at this scale too - 22 BILLION in revenue in a quarter, 12+ BILLION in earnings in a quarter). There are none.

28 Likes

Another thought: Do you really need to invest in AI to achieve success? What are your purpose in learning about and investing in stocks?

If the purpose is simply to achieve a better life, do you really need to focus on AI? I asked myself many times this question.

10 Likes

Thought experiment:
Nvidia begins to lap its hypergrowth quarters which drag on its YoY numbers, both revenue and earnings. We all know it is the same company but it LOOKS much worse than it does right now, and perception rules. The ramp rate of companies spending money on their products slows. They begin to realize the benefits of AI are further off than anticipated so instead of continuing their steep ramp rate, they level off and continue to spend at the existing rate instead, leading to flat growth for Nvidia, which will likely drag on earnings.

This is what I mean by hype cycle. I don’t think the scenario above is too farfetched because it has happened over and over again in the past. Nvidia will continue to be a behemoth but I believe you would agree their growth rate is not sustainable, so then how will the market react when that day comes?

Again this is all just conjecture and I could just be a timid investor missing out of a great opportunity to increase my wealth. Time will tell.

14 Likes

My point is it seems like the market is already/has been wary of that, which is why forward PE is in the 30s - can the company grow earnings by 30% for a decade? There are levers like price increases (the new chips they announced are actually going to decrease their margins) and cost decreases they have not even tapped yet because they haven’t needed to. 30% EPS growth is a fair trade for paying for PE in the 30s.

I think because it’s 7Xed in 18 months it’s easy to look at the STOCK PRICE and say oh obviously that’s hype/overdone and that can’t continue, but there’s a several years runway for the company to grow earnings by at least 30% (they’re doing that per quarter right now, which is +212% a year or tripling their EPS)

Timid? maybe, I’ve also not been brave enough to add to my existing SMCI position or buy NVDA during this run, so you’re not the only one missing out - I’m “guilty” of it too. Just pointing out that the numbers back up the stock growth, it’s not a mere story driving the increase

17 Likes

I agree with you here, at least maybe in our lifetimes. But I do think it will show up in “everyone’s” business in our lifetimes. It is already becoming a very useful B2B service. It’s hard to imagine how that will play out because I assume we don’t have all have intimate knowledge of the day to day operations of various industries, but I bet we could all do a thought exercise to imagine how it will play out in our area of expertise. Here, I’ll start, since it is already affecting my industry in major ways (creative marketing/advertising but you could just call it “consulting”).

My day job is to sift through business, category, consumer and cultural research to put together bespoke marketing plans for companies. There are teams of people (and myself, oftentimes) that gather this and assemble it into insights and recommendations. They get it like 85% of the way there and then I and other senior leaders polish it, make it more creative/bespoke and less boiler-plate, and sell it, more or less.

If I had a great AI developer at my side, I am confident I could employ AI to replace 75% of the staff and give their salary to whatever is enabling the AI workflow, and we’d move 5x faster. The remaining 25% would really only be the ones that I’m training to craft, storytell and sell.

In my industry, I don’t think it’s a matter of if, it’s a matter of when. I just read a document from a strategist where she took us step-by-step through an AI driven strategy for Chanel No. 5 (with her guiding various “pay to play” AI tools). And it was great! People like her who are able to learn this way of working will replace 5 people like me who are too set in old-school ways.

This is happening throughout my company already. What used to take a team of designers now takes 1 designer who knows how to use Midjourney. We have AI-driven processes producing marketing materials for multiple clients where we are again, just in charge of craft. For us, the potential of AI is replacing the entire junior workplace.

There are dangers in creative industries such as mine that thrive on culture and authentic engagement with real people. But we’re probably the exception, not the rule.

23 Likes

I have never seen a technology, not only implemented, but become important so quickly. It’s mind-blowing how I’ve seen it used in the industries I work: podcasts and film. I have always held NVDA because I used it as a more aggressive QQQ: Nvidia plays in every section of tech: AI, Gaming, Autonomous Vehicles, Data Centers…you name it. I added today, at this price, because I think the world is going AI fast, and Nvidia is driving it. Some stocks who are just tangentially using the technology, are getting blown up just by using the word. But some are key. Whether the prices of these key players are stretched at the moment, I can’t imagine them not growing into them in the near future. Nvidia in particular, I regret not adding sooner because I’ve been watching it run, but has anyone seen a company this large grow so rapidly? It’s insane. So I think you need some money in the key AI players, and you just have to pick your spot when you want to jump in. AI is going to be part of daily life sooner rather than later. Just hope it doesn’t murder us all in the process.

23 Likes

AI is almost 70 years old!

The Conference that Started it All

The Logic Theorist was a program designed to mimic the problem solving skills of a human and was funded by Research and Development (RAND) Corporation. It’s considered by many to be the first artificial intelligence program and was presented at the Dartmouth Summer Research Project on Artificial Intelligence (DSRPAI) hosted by John McCarthy and Marvin Minsky in 1956.

What’s new is the neural network implementation and for that to happen there was a need for giant and speedy supercomputers. What is fascinating is how a company dedicated to computer games developed a technology to make games look realistic which has been repurposed to train humanoid robots.

Denny Schlesinger

11 Likes

This is a very interesting thread. I don’t have a lot to contribute, but in particular I find your confidence in Soundhound maybe a bit too enthusiastic. I took a fairly significant position at about $6.20 and sold almost all of it recently at about $8.25.

It’s just a tiny company. Do they have a long term future? Maybe, so far, in my experience tiny companies tend to be high risk. Of course, big companies might also be high risk, so I’m not sure that observation is worth a lot. I kept a very small position just to make sure I’ll continue to watch it pretty closely. I might buy back in.

I guess I’m not 100% sold on voice interface being the future input device. Language tends to be enormously complicated, subject to regional change and frequently ambiguous (I almost got a linguistics degree about 100 years ago. I’ve read works by Whorf, Sapir and several others).

Read Denny’s comment above on modeling the brain. Natural language is intimately related to perception and thought. My wife is Chinese. Her native tongue is Mandarin. Translating Mandarin to English and vice versa is no simple task.

Soundhound does not appear to be doing very sophisticated things so far. The applications they have to offer are all pretty simple and don’t really require intense compute. I just don’t know how far they can take it until they run into a wall.

10 Likes

I tend to agree here, and that is why I am not sure your everyday pedestrian is going to get a lot of use out of things like an AI robot or other tools. If it is not through voice commands, then they would have to learn programming or some other such skill they most likely won’t be up for. I am getting really jaded on the idea that new tech is going to make everything accessible to everyone.

One example was that Android was supposed to be completely open source and everyone can make any app they want. Right, for ONE version of android or whatever skin your phone supplier adds. After that you become in the maintenance game and it’s a slog. I made plenty of apps for my initial smart phones, only to find they didn’t work anymore after a few updates. Then all the really cool small developer apps started to lose support and were never updated for future versions of Android. (So that ENTIRE super smart community of makers seems to have vanished in time.)

I learned the same lesson with 3d printing. Every new printer had some new kind of config set up, different tools and old print designs had to be retweaked to work again.

So, I am probably getting way OT with my own ideas of how far AI stuff is going to go. It has its applications, but it will not be as ubiquitous as the overall AI bubble is making it appear.

(ps - my SOUND position is my most risky one and I agree with keeping it a smaller position.)

5 Likes

Lots of ways to think about what’s going on with AI right now. The OP framed it fairly negatively by calling it ‘hype’, but it’s basically the same question I’ve been asking as well: where are we in the adoption cycle?

Instead of comparing AI to 3D printing and blockchain (two technologies that have yet to find ongoing needs to serve), we could compare to some other adoption cycles:

Note that looking back all the way to the adoption curve for the telephone and electricity, it seems clear that adoption curves have been getting steeper in modern times. The “new new thing” really is coming at us faster and faster.

AI seems to be on a very steep curve right now – but that doesn’t tell us whether the curve is about to flatten or not.

We can say with confidence that companies in just about every industry are either exploring AI’s potential or they are investing heavily in it. Until we see the fruits of those efforts, we won’t really know what the shape of the adoption curve is going to look like beyond the “early adopters”.

Meantime, I would suggest that we all “stay curious”.

ActonUp

25 Likes

if the purpose is to make high returns over the next few years in the market, then the answer to this is resoundingly YES.

if the purpose is become more productive in your every day tasks, then the answer to this is resoundingly YES.

AI is coming for you and the betterment of humanity. Why would you ignore it as an investor?

-muji

36 Likes

That’s only 1 of the 3 reasons:

  1. Power efficiency. But Intel and AMD responded with power efficient chips.
  2. Cost. Intel insisted, and still insists, on making and selling the chips. With ARM, you find your own fabs, which could end up costing less.
  3. Customization. This is big for companies like Apple that want to have special functions or just different trade-offs in the design of the chip.

That said “beat out” is in my mind mostly in terms of popularity. Intel’s model is quite profitable, while ARM pretty much only makes money from licensing. I think if it were such a great business investment, they wouldn’t have sold to Softbank, Softbank wouldn’t have divested it, and they wouldn’t have considered selling themselves to Nvidia (the UK stopped that).

Tulips had/have little to no intrinsic value. 3D printing has its uses, but it’s not going to replace lower-cost volume manufacturing.

OTOH, AI is going to affect everything.

AI will end up in everyone’s house via domestic applicances. My vacuum cleaner already has AI in it that controls the strength of the suction to conserve battery without hurting results. AI will be in everyone’s phone, in everyone’s car, and involved in almost all business decisions. You’ll use it to write resumes and to inquire about job openings. Young-ins will use it write love texts/emails. It’ll be everywhere. Every Where.

This is short-sighted. Teslas already drive on roads they haven’t been on before. Yeah, not perfectly, but that’s just a “yet” qualifier.

Same for recommendation alghorithms - whatever FB is doing today is much less sophisticated that what will come in just a couple years.

Was mobile computing a hype cycle? If you invested in Apple in 2007/2008 because of the hype, would you be happy or disappointed today? If I had told you in 2008 that everyone would have much larger than Nokia tiny phones that were actually worse phones, would you have believed me? Microsoft’s Steve Balmer famously laughed at the iPhone. Mobile phones became mobile computing, led to tablets and watches, and now all sorts of SAAS that make money for Apple and others. We’re now in a world where everything is connected, companies like Samara are still growing their connectivity businesses, and DVDs are almost as quaint as vinyl.

As for the hype cycle, the operative phrse above is “…benefits of AI are further off than anticipated…” Which way are you going to bet money on that? If you really think AI is going to be like 3D printing (which thanks to Bambu Labs is only now getting to be easy to use) and not like mobile computing, then sure, avoid AI companies.

As for Nvidia, we’ve talked about this a few times before. I don’t see Jensen Huang being like Cisco’s John Chambers, and not moving his company’s offerings as the world advances. Arista was founded in 2004, by people who left Cisco because it wasn’t moving with the times - and Cisco is still struggling to catch up on SDN (Software Defined Networking).

Huang isn’t making those kinds of mistakes. He almost got away with buying ARM, but even so Nvidia arguably makes the fastest/most efficient ARM chip for data centers, which they pair with their best in the world AI chips. Nvidia’s got software offerings (making $1B/year in revenue already), rental offerings, and cloud offerings.

Mr. Market is already with you there. As has been pointed out, Nvidia’s not priced at 100x earnings, and their PEG is still more than reasonable. But more importantly, how far off is this comeuppance? By the time NVDA’s stock gets punished, I’ll have 10X’d my investment. Look at TSLA - after its end of 2021 hype it’s down like 60%. And while I’ve taken profits along the way, I still have some shares that I bought which are at least 70X their buy price.

36 Likes

I have thoroughly learned two languages other than my mother tongue, one of which (Japanese) is about as different from English as can be imagined, and back in the day it was taken as self evident that only humans could ever tackle translation. That received wisdom has been long since turned on its head, the result of increasingly powerful compute, and llms.

Speech is “the” mode of language. It in fact is the very definition of it. Written communication is a johnny come lately, linguistically, and among the youngest generation it may even be withering.

I think it is only a matter of time before AI can elegantly handle voice, and given how quickly we are seeing progress, that time is likely short.

17 Likes

Interesting observation.

Not all that long ago, when my wife and I were just beginning a long distance email relationship I would write in English (my Mandarin is marginal and pinyin/hanzi even less). I would then run it through a software translation to Mandarin. I would then take the Mandarin output and translate it back into English. After that I would try to fix all the things that didn’t come out as intended and repeat the process. I would iterate as necessary until I got a reasonably accurate English translation from the Mandarin.

Might I have accomplished the same with a voice input? I doubt it. Writing demands an organizational level driven by grammar which is seldom present in speech. Writing requires discipline not present in speech. I wrote countless technical documents during my career at Boeing. It was a time consuming activity. I venture it would have taken longer if I just “talked” the content.

Will written language die out? I really doubt it. Verbal contracts generally don’t hold up too well in court. The Chinese contracted with Russian engineers for the construction for part of the three gorges dam construction project. All the contracts were written in English.

STE100 (Simplified Technical English) is a worldwide standard for aerospace products maintenance manuals. Doesn’t make any difference if the plane was built by Boeing, Airbus, Embraer, Comac and I imagine all the smaller business jet makers. The manuals are written with the controlled vocabulary and standard grammar as defined by STE100.

So I see a limited scope of applications for Soundhound. I’m not saying that company can’t grow and be successful. Even a limited scope provides a pretty broad opportunity. But I’m not as sanguine as many others regarding the future of this company.

9 Likes

You need to consider that AI models are being layered together. LLMs are giving every other AI a method of human communication. You can then add a speech-to-text model over that LLM to add voice control, and text-to-speech to have the LLM talk aloud.

That means you can now use typing or voice as an interface into the computer. All the other cutting AI systems being developed (like autonomous vehicles, or the coming wave of humanoid robots being talked about in another thread) can now use those AI models above to fully communicate with humans bidirectionally.

Of course, there is a ton of market jubilance over certain names showing the how hyped it all is (ahem SMCI), but this is just the beginning of the AI Era. We’re still out the infra buildout stage. There is a LOT more to come in future directions for AI, as all new app layers and various directions in AI solidify into products & get deployed from here. This wave of LLMs is just the start.

-muji

34 Likes

Smorgasbord, GREAT reply. Thank you for the effort of rebuttals in a nice way!

12 Likes

At some point in time, you’re probably right. But to put things in perspective, I’m pretty old, I think it’s unlikely that I’ll see much of this during my life (despite the astonishing pace of technology). That’s not to say that I can’t profit from it now if I can only figure out where to invest most productively.

So I guess I’m not fully sold on the bidirectional speech interface you just alluded to being the best investment at this time. Let alone SOUN being the company that will deliver on it. It’s a tiny, immature company. Of course if it does turn out that they do deliver, then now is the ideal time to get on board.

But I’ll add (actually, reiterate) spoken language and written language are two quite different things. You, as a blog writer, should be fully aware of this. Writing (usually) involves more thought than speaking. Anyone who has ever tried to explain something somewhat complicated understands the practice of rewriting the same things over again for clarity and precision, sometimes more than once. But once you say something, it’s out there, you can immediately try and assert, “oh, that’s not what I really meant” or whatever, but you can’t truly unsay it. Every spoken word and sentence is a bell that can’t be unrung. Every non-sequitur, dangling participle, noun/verb mismatch, ambiguous reference and other sins of speech is permanent.

I’m not in the least ambivalent about the promise of AI. But as you noted, the infrastructure is just beginning to be developed. There are a few crude examples that have been presented, but much needs to be done before any of the fruit fully ripens.

7 Likes