Hey Everyone,
So on October 12th, Airstreet Capital put out their widely celebrated State of AI Report 2023.
In my first analysis of the report, I covered up to Research. Now it’s time to cover the rest. You can read the first overview here.
🏆Sponsored⭐
AI: What Consumers Say
Will consumers reward ethical data use in the age of AI? We completed a first-of-its-kind national survey to examine sentiment around artificial intelligence and the personal data that powers it.
This will include many of my own comments and analysis on the different points of the report.
For more deep dives you can support the channel.
🔸Now in its sixth year, the State of AI Report 2023 is reviewed by leading AI practioners in industry and research.
🔹It considers the following key dimensions, including a new Safety section:
Research: Technology breakthroughs and their capabilities.
Industry: Areas of commercial application for AI and its business impact.
Politics: Regulation of AI, its economic implications and the evolving geopolitics of AI.
Safety: Identifying and mitigating catastrophic risks that highly-capable future AI systems could pose to us.
Predictions: What we believe will happen and a performance review to keep us honest.
We’ve covered the Research section already, so we have a a lot to summarize in this second installment. Not all of you have the time to go through those 160+ slides, so I hope to cover some of the important parts.
Which takes us to slide 69.
Industry
The GPU demand caused by the Generative A.I. craze made Nvidia a big winner. Q2 ‘23 data center revenue was a record $10.32B, up 141% from Q1 ‘23 and up 171% from a year ago. The Nvidia stock is thus up 222% so far this year in 2023.
This means Nvidia’s stock price has a P/E ratio of 111. Which is likely not sustainable but illustrates the 2023 frenzy around A.I. chips. Microsoft and OpenAI have been working on their own A.I. chips, and Saudi Arabia is highly likely to be collaborating with China, even as the U.S. prepares for more detailed A.I. chips sanctions on China.
The demand for Nvidia’s A.I. Chips in 2023 means they have a monopoly (between 70 and 80%) on the demand and need to boost their supply for 2024.
CoreWeave, Lambda, and Crusoe Cloud, three selected NVIDIA partners that build and run GPU datacenters, together have tens of thousands of GPUs in their fleet. Demand outstripped supply by many multiples in 2023.
Well funded startups like OpenAI, Anthropic, Inflection, Imbue and Cohere thus have an advantage of getting ahead in line for these A.I. chips by Nvidia. Inflection also also been helped out by Microsoft, Bill Gates and Nvidia themselves. Until the costs go down, a lot of their outrageous funding just goes to the compute and hardware costs of all the GPUs and expensive salaries of A.I. scientists and engineers.
In other words, things would have been moving even faster in 2023, had this bottleneck of A.I. chips not occured.
AI Chips are the new oil in the era of A.I. Supremacy?
I like how Nathan goes on to make lavish statements to this effect. China has found workarounds to access Nvidia’s H100 GPUs.
Saudi Arabia’s King Abdullah University of Science and Technology (Kaust) has allegedly purchased >3,000 H100s to build a supercomputer, Shaheen III, that should be operational by end of 2023. Its LLM-focused researchers are primarily Chinese nationals that cannot access the US because their universities are restricted.
You could make rankings just based on H100 Clusters. The number of large-scale NVIDIA A100 GPU clusters has grown since last year, particularly at Tesla and Stability A.I., as well as new clusters at Hugging Face. Nvidia actually has become involved in funding many of the key startups in the space. This means the A.I. LLM era has made Nvidia part of the ‘magnificent seven’ BigTech leaders of monopoly capitalism. Typically when you enter the trillion dollar club, this is what it signifies at least relative to 2023 market cap valuations.
You need well over ten thousand H100s to power Generative A.I. at a world-class level in your clusters. This is why companies like Microsoft, Google, Apple, Meta and Tesla have the ability to do things other companies won’t be able to do. It’s not just A.I. chips but the supercomputers that they are able to build.
The New A.I. Chip Arms Race
H100 infrastructure for large-scale model training. As of writing, Google and Inflection are not yet at full scale and we understand others including OpenAI, Anthropic, Meta, Character.ai, Adept, Imbue, and more have significant capacity.
Generative A.I. is not exactly a level playing field here. To keep up with OpenAI and Anthropic, A.I. startups will need major cloud backers and multi billion dollars in funding. Only to likely get acquired by a magnificent seven company in the end anyways.
To give you an idea on how prefered Nvidia chips are in A.I. training:
NVIDIA chips are used 19x more in AI research papers than all alternative chips combined.
It appears that Nvidia’s H100 can last a good 5-7 years which gives Nvidia plenty of time to come up with vastly better upgrades. Nvidia’s advantage here is immense compared to its competitors which explains why some still consider the stock cheap today even after its sizeable growth in 2023 thus far. Given the future of A.I. chips, Nvidia’s $1.14 Trillion market cap today might just be the beginning of its growth phase.
Highlights
Cerebras, creators of the largest AI chip in the world, engaged in several open source model training and dataset creation projects, which helped it gain traction versus its competitors with researchers.
Tesla marches towards a Top-5 largest compute cluster for AI in the world.
Cerebras
Graphcore
Cambricon
Habana
How much will Nvidia’s Demand Grow in 2024?
It is also rumored that NVIDIA is to ship 1.5M and 2M H100s in 2024, up from the 500,000 expected this year.
So a 3X growth in sales could be significant for Nvidia’s growth as a company.
Who are the Big Spenders?
Microsoft
Baidu
Tencent & Meta
A bit surprised to see Nathan make this claims, which is a widely popularized myth.
OpenAI’s ChatGPT is one of the fastest growing internet products
Unfortunately one year later, this is no longer true. Sort of what you expect from a Venture Capital fund though. No going down a bit in revenue due to ChatGPT doesn’t mean it was disrupted by OpenAI!
GitHub CoPilot drives significant productivity gains for developers
Microsoft’s GitHub Copilot is actually losing money, at quite an astonishing rate, but glad to hear that it’s helping developers be more productive (or deskilling them).
Studies by the companies involved cannot be considered real data! In June 2023, GitHub reported data from 934,533 CoPilot users. Interestingly, productivity dips a little bit before significantly increasing as Copilot users get acquainted with the tool, and the less experienced users are the ones who benefit the most (~32% productivity gain).
In fact, OpenAI’s ChatGPT likely took marketshare from GitHub Copilot, which was a pretty poor conflict of interest for Microsoft in the first place.
GenAI has a User Retention Problem
Sequoia capital data:
This means Generative A.I. products don’t actually provide enough value to their core users to reduce churn or to gain much time in product of DAUs and MAUs compared to other applications.
ChatGPT doesn’t have the retention rates of a product that has found product-market fit even after one year of development. Other A.I. startups fair even more poorly.
You can refer to the source here.
If the entire hyped industry doesn’t provide real consumer value, one has to assume that the unlimited funding we saw in 2023, might stop flowing quite the same way in 2024 or perhaps as late as 2025.
The Ratio of DAUs to MAUs is Broken for Generative A.I.
Compared to global consumer apps, the ratio of AI-first companies doesn’t meet the requirements for being or becoming mass consumer apps. The ratio of daily active users to monthly active users is too low, the value prop is too weak.
ChatGPT has a ratio of 14% which in my opinion, is unredeemable, whereas Charater.AI is actually showing far higher product-market fit, mostly with GenZ.
Nathan did not include this slide in his presentation. And for good reason!
Several slides in the State of AI 2023 are wasted on comparing his former predictions with the state of 2023, which is still pretty nascent.
From labels to preferences
Highlights:
As instruction fine-tuning and RLHF became the default method to fine-tune and align language models, companies offering labeling services like Scale AI and Surge HQ stand to register exceptional growth from the exploding popularity of LLMs.
Both companies bolster an impressive list of customers, from AI startups to large corporate clients to leading labs in LLM research. Scale AI was last valued at $7.3B back in 2021, pre-Stable Diffusion and the ChatGPT frenzy.
Scale AI was last valued at $7.3B back in 2021, pre-Stable Diffusion and the ChatGPT frenzy.
Surge AI was founded by Edwin Chen. Its Series A was back in July, 2020. So it’s likely due for more funding soon.
Open source AI’s Incredible Momentum
While ChatGPT hype dominated the first half of 2023, things are pretty different now. You might say that closed AI has lost out in enthusiasm of the community to open-source AI projects and LLMs.
In many cases these are freemium models and not technically Open-source, but let’s not get hung up on semantics.
Companies like Hugging Face give us some visibility here.
Hugging Face, the now 7-year old company that has firmly become the town hall for open source AI, is seeing significant momentum as the community vies to keep AI models and datasets accessible to all. Over 1,300 models have been submitted to their Open LLM Leaderboard in a few months and >600 million model downloads in August 2023 alone. These models are exposed on Spaces as web applications built with tools such as Gradio or Streamlit, enabling broader accessibility and rapid prototyping. Monthly active Gradio users has grown 5x from 120k (Jan ‘23) to 580k (Aug ‘23).
Big Pharma is going Big on Generative A.I.
Few companies stand so much to gain as pharma companies going after AI in drug development.
mRNA vaccine leader, BioNTech acquired AI company InstaDeep for €500M, while Sanofi goes “all in” on AI, Merck enters into new deals with AI-first drug company, Exscientia, worth up to $674M and AstraZeneca partners with Verge Genomics in a deal worth up to $840M.
Nathan has one slide on this, when it easily could have been five. I don’t think it’s his area of expertise.
Google Deepmind V2 is Here
Trying to follow the history of A.I. through the lens of DeepMind UK, that was acquired by Google has been one heck of a ride. Even as Google was threatened by Microsoft and OpenAI likely for nothing.
It’s still unclear if Bard or even Gemini will amount to anything. [Gemini isn’t even out yet]
The Transformers Mafia
The young scientists working at Google at the time of the 2017 paper that made Transformers big, have all left Alphabet, and raised billions. You might call them the OpenAI Mafia as well.
NVIDIA’s NVentures and Andreessen Horowitz gave Inceptive AI funding to the tune of $100 million. It’s one of the less known startups of the bunch. Founded by former Google AI researcher Jakob Uszkoreit, Inceptive is developing an AI platform that can design unique mRNA sequences.
This mafia of startup founders tainted by their affiliations with the magnificent seven means Generative A.I. development is actually very concentrated in the hands of a few venture capitalists and their favored A.I. scientists. Few understand the implications of this with regards to A.I. regulation.
Having a few people, funders or backers controlling the entire movement is obviously not good for innovation.
GenAI is a Just a Sliver of Excitement – “GenAI” is the new “new” thing: But Overall, AI investments are stable vs. 2022
Generative A.I. is like the 🍒 cherry on the top of AI funding, not as significant as one might think.
The Problem with 2023 Valuations
Notably, 50% of the S&P 500 gains in 2023 were driven by “The Magnificent Seven”: Apple, Microsoft, NVIDIA, Alphabet, Meta, Tesla and Amazon as key drivers and beneficiaries of AI acceleration.
Saying “AI” multiple times in your earnings call was enough to get shareholders excited.
As you might recall in 2022, the Nasdaq-100 finished the fourth quarter of 2022 with a full-year loss of 32.4%.
Generative A.I. hype has helped it rebound in 2023, and then some:
The NASDAQ 100 is up 39.68% to date as of October 17th, 2023 year to date.
The conclusion has to be that BigTech used Generative A.I. concentrated in just a few of its former employees to push shareholder value of their companies back up. But is it sustainable?
OpenAI revenue of $1.3 billion annualised rate is mostly just selling GPT-4’s API to unsuspecting Enterprise clients. Even as OpenAI is likely hugely unprofitable, yet allowing employees internally to share shares at a $90 billion valuation. It’s like a pyramid scheme.
Generative A.I. has also been a vehicle to get more money flowing back to the United States, even as economies like China and Germany really are falterting in 2023.
US AI companies absorb 70% of global private capital in 2023, up from 55% in 2022
As we know the same cities in the U.S. get the bulk of all of this funding. Although layoffs at BigTech were massive in 2022 and for some still continue to this day. Generative A.I. thus appears to be a scheme that makes Silicon Valley even more powerful and more centralized.
As China’s economy has slowed dramatically with a real-estate, debt and youth-unemployment crisis and Europe is but a shadow of what it once was, the U.S. can super-charge Generative A.I. to benefit their Cloud, digital advertising, and software services leaders.
The United States is in full control over the closed AI Generative A.I. movement thus far.
The US continues to lead by number of AI unicorns, followed by China and the UK
The from 2022 continues: the US grows its unicorn count to 315 from 292 and total enterprise value to $5.9T from $4.6T. The UK adds 3 more unicorns but sees cumulative enterprise value regress to $155B from $207B.
2023 has been the Definitive Year of Generative A.I. Funding
It should be noted that if you take out OpenAI and Anthropic as outliers, it’s not so incredible as it looks. For the most part these are people who used to work in BigTech and more will follow.
Anthopric is likely to get $2 more billion from Google and up to $2.5 Billion more from Amazon. It seems almost inevitable that the Cloud companies will own Generative A.I.’s best apps and capabilities. This even as the magnificent seven more more into healthcare, finance, banking, education, automation and robotics.
2023 Saw a Large Preponderance of Mega Deals (over $100 million)
Considering a macro environment of higher interest rates, a majority of these well-funded startups )(in pink) won’t make it – meaning the 2023 activities are likely setting us up for a hype bubble popping sometime in the not so distant future.
Generative A.I. might be the biggest tech bubble in history. Being compared to the internet or mobile is not a great look. What a spectacle.
BigTech were among Biggest Funders of Generative A.I. Boom of 2023
Nvidia, Microsoft and Salesforce in particular were very active in 2023 funding A.I. startups.
Compute and talent isn’t just expensive in Generative A.I., it’s likely exaggerated by the hype.
GenAI companies raised 33% larger Seeds and 130% larger As than all startups in 2023.
Typically startups that get over-funded don’t tend to fare that well as we learned with the Softbank Vision Funds’ failure rate in recent years.
In facto according to a recent report by Accel, U.S. tech giants added $2.4 trillion to their market capitalizations in a year defined by the hype around generative artificial intelligence. Read the entire report.
Politics
This takes us to slide #121. I will try not to dwell on this section.
It’s hard to have faith in Government specified rules for A.I. specific legislative frameworks. This is because as geopolitical instability occurs, BigTech is working more closely with National Defense and Government sectors. If anything, monopoly capitalism winners should get more entrenched with Government and the Pentagon than ever in such a macro geopolitical environment for the foreseeable future. Lobbyists and former CEOs of these companies therefore can steer A.I. regulation that favors these corporations.
Highlights
So far, both the UK and India have stressed the economic and social upside of AI, with the March 2023 white paper and a parliamentary response from India’s digital minister arguing that any current risks could be absorbed by current sectoral regulations and privacy legislation.
The UK also secured a special agreement with Google DeepMind, Anthropic, and OpenAI to gain early access to their most advanced frontier models to improve their understanding of risk.
The EU and China are leading the pack in passing new, AI-specific legislation, with especially stringent measures around foundation models.
The EU’s AI Act is entering its closing legislative stages in the coming months. The Parliament’s current draft has added regulations around foundation models and general purpose AI systems (which are stipulated separately).
China brought in specific legislation on recommender systems, alongside generative AI regulations. This updated previous ‘deep synthesis’ regulation that required AI-generated content to be labelled, protections against misuse, barred anonymous accounts using services, and included censorship requirements.
The US is unlikely to pass a federal AI law anytime soon and in some respects is pursuing a UK-style approach, with an emphasis on voluntary commitments (e.g. the July White House agreement) and research to establish what constitutes good practice (e.g. the National Institute of Standards and Technology’s AI Risk Management Framework).
Canada is attempting a slimmed down version of the EU AI Act, banning certain and applications and regulating others.
If the majority of A.I. companies and big players are in the U.S., how other countries regulate Generative A.I. and the absence of serious regulation in the U.S., is a huge red flag.
The UK is planning to host a summit themed around safety and governance in November 2023, involving major democracies and China, as it attempts to position itself as the world’s leading safety research hub.
Anthropic, Google, OpenAI, and Microsoft launched the Frontier Model Forum – a body designed to promote the responsible development of frontier models and to share knowledge with policymakers.
The U.S. wants to be free of specific Generative A.I. regulations that might hurt it in its A.I. arms race with China. However, this might end up being dangerous for the rest of the world and civilization’s future.
The U.S. A.I. chip bans on China means Taiwan’s “Silicon Shield” will be greatly diminished as a deterrent from China invading the Island. TSMC is building in Phoenix and will build in Germany. Many other countries are joining the U.S. in these bans.
Many analysts believe China is preparing for war or a likely invasion of Taiwan by as early as 2025 and likely before 2028.
Japan toughened its export limits to include the equipment make less-advanced chips. The Netherlands has widened its restrictions on the export of the deep ultraviolet lithography machines.
The semiconductor is a complicated global supply chain, and sanctions and trade bans don’t make a lot of sense. Huawei managed to still release a new phone, even with significant sanctions in place against it. This has impacted Apple sales in China in recent weeks.
China’s economic implosion likely means it needs to step up its war games, before it’s no longer feasible to do so. China is generally seen as about 9 months behind the U.S. in Generative A.I. advances.
U.S. National Defense Funding Rises
Funding for US defense startups hit $2.4B last year, more than 100x the European total, but the number of companies able to win consistent, sustained work remains small.
Europe appears all but left behind. For example, Anduril’s Series E is greater than all British defense tech investment between 2013-2022 combined, while Helsing’s €209M Series B is the only significant fundraise on the continent.
Europe’s inability to keep up in Generative A.I. startups, cutting edge National Defense startups or Quantum computing endeavors is pretty alarming.
According to Nathan, European LPs largely aren’t reversing their aversion to defense investment, meaning that supranational institutions are stepping in to fill the gap.
Ukraine and Israel act as labs for AI warfare
Israeli’s National Defense startups are thriving. And the U.S. and China are watching closely. Drone asymmetric warfare has become the norm in the Ukraine. I’ve seen a huge uptick of U.S. weaponization of A.I. and robotics startups of late in 2023.
But it’s pushing the world to automate its military systems at a pace that could become fairly dangerous in the 21st century.
After successful trials in 2022, the Ukrainian Armed Forces fully authorised the use of Delta in February. Delta is a cloud-based situational awareness system that integrates data in real-time from different sensors, satellites and drones, along with intelligence or images taken from those on the ground.
Who needs more soldiers when we will have more complex A.I. systems controlling killer drones and swarm technologies? Be careful what you wish for.
Are the ‘culture wars’ coming to AI?
With Presidential elections coming to Taiwan and the U.S. in 2024, A.I.’s impact on democracy is going to get potentially very messy.
But Hollywood writer strikes and other events show a weird disconnect.
ChatGPT has become a flashpoint in a series of heated cultural debates, largely in the US, with particularly conservatives sharing screenshots to allege bias in ChatGPT’s training and fine-tuning.
Everyone from artists, writers to famous published authors to cloned actors are protesting, entering lawsuits and debating how A.I. might impact what they do. AI as plagiarism and cheating normalizers and the risks to senior citizens via phishing is not exactly protected.
Nathan notes the weird alternatives to OpenAI on the political spectrum. Following long-standing complaints about OpenAI’s “political correctness”, Elon Musk launched xAI, a startup focused on trying “to understand the true nature of the universe”. In a Twitter Spaces following the launch, Musk emphasised that “our AI can give answers that people may find controversial even though they are actually true”. Little is yet known about xAI’s work. Though we note that its X account has over half a million followers.
Everything from copyright to how Generative A.I. might impact elections is up in the air. Clearly LLMs are trained on our data without our consent and this is becoming normalized by BigTech. Scrapped data, reddit, what we say on Meta’s apps, everything is game!
Key Takeaways
Key themes in the 2023 Report include::
GPT-4 is the master of all it surveys (for now), beating every other LLM on both classic benchmarks and exams designed to evaluate humans, validating the power of proprietary architectures and reinforcement learning from human feedback.
Efforts are growing to try to clone or surpass proprietary performance, through smaller models, better datasets, and longer context. These could gain new urgency, amid concerns that human-generated data may only be able to sustain AI scaling trends for a few more years.
LLMs and diffusion models continue to drive real-world breakthroughs, especially in the life sciences, with meaningful steps forward in both molecular biology and drug discovery.
Compute is the new oil, with NVIDIA printing record earnings and startups wielding their GPUs as a competitive edge. As the US tightens its restrictions on trade restrictions on China and mobilizes its allies in the chip wars, NVIDIA, Intel, and AMD have started to sell export-control proof chips at scale.
GenAI saves the VC world, as amid a slump in tech valuations, AI startups focused on generative AI applications (including video, text, and coding), raised over $18 billion from VC and corporate investors.
The safety debate has exploded into the mainstream, prompting action from governments and regulators around the world. However, this flurry of activity conceals profound divisions within the AI community and a lack of concrete progress towards global governance, as governments around the world pursue conflicting approaches.
Challenges mount in evaluating state of the art models, as standard LLMs often struggle with robustness. Considering the stakes, as “vibes-based” approach isn’t good enough.
Since I’ve been covering A.I. news and reading A.I. papers a lot in the past year, I did not find the State of AI report 2023 very comprehensive or novel as I had remembered from past years. However, I hope you learn something new from it.
Every Venture Capital fund will have their pet perspectives and things they like to cover and Nathan and his team are no exception to this.
Hot Industries and Areas of A.I. Funding
Robotics
Enterprise software
Health
Transportation
Cybersecurity
FinTech
Nvidia’s AI Land Grab by Funding
Not only did Nvidia become a huge monopoly for A.I. chips in 2023, but invested very strategically in multiple areas including Foundational models, drug discovery, GPU providers and Cloud specific solutions.
Nvidia, not OpenAI is the real story of 2023 without a doubt as the key theme of Generative A.I.’s hype cycle hitting its peak around July of 2023.
More detailed A.I. chip sanctions will hurt Nvidia’s sales in China.
The tighter controls will target Nvidia’s A800 and H800 chips, a senior US official said, which the American firm created for export to China — the world’s largest market for chips — after the Biden administration introduced its initial restrictions last October. – Bloomberg
Who Nvidia Funded?
Nvidia funded many of the most important A.I. startups outside of OpenAI and Anthorpic including:
Inflection AI
Cohere
Hugging Face
Runway
A21 Labs (Israel)
Outside of Microsoft, Google and Amazon, Nvidia is the most diversified backer of Generative A.I. related startups of the magnificent seven thus far with much smarter and strategic picks that can make Nvidia even more valuable in the future.
Series A Rounds become Bloated in 2023 for Generative A.I. Startups
The bloated nature of Series A rounds for Generative A.I. startups is likely not sustainable as many of these startups will struggle to make revenue. Even though compute is expensive and talent is competitive, the bloated nature of these rounds means they are for the most part likely dead-ends.
Safety
I’m going to gloss over this section, as 2023 is very immature when it comes to any kind of serious A.I. regulation or rules.
Nathan attempts to summarize all that occured in 2023 in the media, but it doesn’t arrive at any real conclusion.
The final twenty or so slides try to summarize safety concerns. Given that the U.S. had no intention of seriously regulating A.I. and lobbying by the various companies involved, it mostly seems like empty public relations, and political show at this point by various organizations and personalities.
The world is a more dangerous place with ChatGPT and all of these other A.I. tools, including serious cybersecurity and phishing concerns. Everything from copyright violations, to not getting consent from consumers, artists or authors for training their LLMs means even to debate safety at this point or set up teams to deal with ASI, is pure marketing and deception.
Most people have no idea about x-risk or A.I. ethics but understand that A.I. alignment is important especially around privacy. Unfortunately OpenAI did not do its due diligence in this regard when it raced to put out ChatGPT.
The idea that labs are building their own mitigation against existential A.I. risks in 2023 is beyond pretentious. The fact that a lot of the Open-source models aren’t even really open-source is another point of contention.
The ease of jailbreaking AI models or getting them to hallucinate toxic content means the majority of products produced by Generative A.I. in 2023 aren’t even safe or at an adequate level of trust & safety. This signals to me that they were released prematurely. Yet there is no regulatory body in the U.S. that seems to take this into account or have penalties for it.
Even with RLHF there are fundamental challenges:
In truth, Generative A.I. products like ChatGPT should never have gone live in 2022, nor likely in 2023 until basic trust & safety was dealt with. Unfortunately that’s not the world we live in and in a centralized ecosystem where Silicon Valley and Venture Capital norms still dominate, we will continue to use LLMs to invent more and more dangerous technologies let loose into the wild until potentially major accidents will occur.
The State of AI report 2023 tries to be everything for everyone, and in doing so, somehow falls flat this year for me. While in years past I was amazed by its depth, perhaps because so much occured in 2023 in A.I., this year it just felt like a selective summary of past events that left me feeling Venture Capital funds are far from experts in what they fund.
While most American VCs over-exaggerate Generative A.I. for the profit motive, at least Air Street capital makes an attempt to be objective. If you knew nothing about Generative A.I., the report might be a good primer. The majority of the infographics were just pulled from papers and other available sources.
Many analysts believe the hype for Generative A.I. might fall flat sometime in 2024. Generative AI has ultimately become “overhyped” and smaller developers of the tech will face challenges as it becomes “too expensive” to run, said Ben Wood, chief analyst at CCS Insight.
Even OpenAI likely has mounting costs and even $1.3 Billion ARR might mean it won’t become profitable for many years, if ever. So it like Anthropic will just have to keep getting more funding. While Character.AI has gotten some traction with younger GenZ consumers, not many apps have. For a technology that was even compared to a General Purpose Technology earlier in 2023, that seems somewhat surprising.
How many of a16z’s list of trending apps in Generative A.I. will even still be around in a few years?
It’s mostly LLMs adopted by Enterprise customers that seem to be the bread and butter of this movement. With a lot of companies doing it themselves with Open-source LLMs. Outside of coding, a lot of this technology is immature to actually boost productivity, revenue or create radically better customer experiences.
Industry leaders like Eleven Labs might survive, but many of these companies and their apps will fade pretty quickly. Yet there are thousands of A.I. tools with tool aggregators even referencing each other. It means there is a great deal of fluff and the early copilots don’t have much utility and aren’t well-liked by users including Bard, Bing and the rest.
When even ChatGPT only has 14% who are DAUs compared to MAUs you get an idea of how little traction even the so-called leaders of these products actually have in the real world with actual customers.
Many in the stock market say that Generative A.I. represents a secular bull-market trend. While that seemed probable in 2023, I’m afraid this hype cycle won’t age well into later 2024 and 2025. I concur with the camp that admits that the buzzy generative artificial intelligence space is due something of a reality check next year or in early 2025.
For that reason specific domains like A.I in drug discovery and robotics might be as important as many of the advances specific to Generative A.I. As A.I. chips improve and LLMs and SLMs evolve with more multi-modal capabilities, more interesting things will become possible.
I’m afraid Venture Capital funds just aren’t the ideal storytellers or historians/archivists when it comes to a movement they themselves fund and back. Their own incentives skew their perspective or perhaps they actually drink the Kool-Aid in the adrenaline rush of the early days. Already by October, 2023 it appears that Microsoft’s investment in OpenAI has been a failure, nor is Microsoft set to be leaders in A.I. by throwing $10 Billion up to $13 Billion at least already at this team. But at least Microsoft have a few more subscriptions (Copilot 365, GitHub Copilot, etc…) they can bundle into their software services in an anti-competitive way.
With all the betting and gambling in the space, time will show how few of them ever actually become profitable. A number of categories already seem to have early leaders or leading duopolies, which doesn’t bode well for a real and competitive innovative ecosystem. When who can fundraise the best, dictates the early winners.
Read More in AI Supremacy
