Newsletter 19: Predictions are risky business but one quarter into 2024, mine are doing fine.

The promised quarterly update on 2024 predictions.

Apr 10, 2024

On January 3, 2024, I published my 2024 predictions on LinkedIN. If you haven't read them, here is a link. As I had written then, I focussed on talent, data, productization or vertical focus and infrastructure to make my bets. I had observed that these were directly affected by capital and that capital deployment had already started in a very big way.

To get to these narrow set of constraints, i used a framework I developed over the Thanksgiving break as mentioned in essay #10, “Navigating the AI Labyrinth: Silicon Valley's Modern Odyssey”.

In general I was on the right path, a few are definitely validated, like boulders rolling down a hill, these will only gain in momentum. The others are still developing and maybe take longer than I had expected. None seem to have been refuted so far.

So let's get into it.

Prediction 1 - The AI fight started in 2023 will get entrenched - On one side we will have a subset of Big Tech who have influence or indirect control over “Proprietary AI” and on the other sides will be “Open AI” fueled by aspirants, venture capital and those in Big Tech who missed the first wave. In 2024, the first movers will maintain their advantage primarily because of early talent sequestration.

Q1 Outlook - Trending as predicted.

The most clear sign that this prediction is holding true, at least as far as influence or indirect control of “Proprietary AI” via talent sequestration is concerned, is the announcement that Microsoft hired AI veteran Mustafa Suleyman last month to lead its AI efforts. About 70 Inflection AI employees joined the cofounder in jumping ship to bat for Satya Nadella's team and former Inflection and DeepMind scientist Jordan Hoffmann is leading a new AI hub for them in London. In another intriguing move which has become Microsoft’s signature tactic, this is not an acquisition of Inflection, which recently secured a staggering $1.3 billion in funding just last year and was a very high profile AI startup through 2023. This builds upon Microsoft's deep partnership with OpenAI as I have written about in this series earlier.

Meanwhile, Amazon and Google have invested in Anthropic, a generative AI startup, and Amazon has a minority position in the company. Amazon invested $1.25 billion in Anthropic in September 2022, and then an additional $2.75 billion in March 2024, for a total of $4 billion. Google also has a 10% stake in Anthropic.

There are many other such moves afoot, for example, Logan Kilpatrick, who left OpenAI last month, announced that he's joining Google to "lead product for AI Studio and support the Gemini API." Tesla is reportedly raising compensation for its artificial intelligence engineers in a bid to ward off poaching from the likes of OpenAI. Elon Musk said that the competition for AI engineers “is the craziest talent war I’ve ever seen.” He should know, Tesla machine-learning scientist Ethan Knight left the automaker to join Musk’s AI startup, xAI. Musk said, “Ethan was going to join OpenAI, so it was either xAI or them.”

On the “Open Source AI” side, the vanguard is also galvanizing. Elon is not only acting as an AI gadfly by suing OpenAI, he also said last month that his xAI startup will open-source its chatbot Grok, which hasn’t captured as much interest as OpenAI’s ChatGPT or Anthropic’s Claude.

Last year Meta released its Llama 2 open-source model, intensifying the competition between private and open-source models. Interestingly it was made available on Microsoft's cloud platform also. Google released its Gemma open-source models in February, though it also sells access to its more powerful Gemini offerings.

Plenty of AI startups are betting on open source, too, with some of the most well-funded including Mistral AI, Hugging Face, Runway ML, Together AI, Writer, Cerebras and Databricks. In the case of Mistral, the Microsoft intrigue continued when the company invested €15 million ($16 million) into Mistral and announced it would bring Mistral’s newest AI model, Mistral Large, to Azure.

In this fight Microsoft seems to be reaching out and ensuring it is covering all bases. Maybe it is for access and connection to talent, maybe it is a way to distribute its risk. Definitely more to come.

Prediction 2 - AI will accelerate the already torrid Data Center growth further driving energy demand. Given the dependence on specialized chips and requirement for dedicated use for model training 2024 will see growth in Data Center usage dedicated to AI use by hyper scalers as well as other companies. This will be an incremental demand in 2024 on top of the forecasted demand for data centers by hyperscalers and others leading into 2023. REITs that specialize in this class of real estate should see huge upside.

Q2 Outlook - Done deal. This has happened and will gain pace.

Look no further than Microsoft (again!) and OpenAI working on plans for a data center project that could cost as much as $100 billion and include an artificial intelligence supercomputer called "Stargate" set to launch in 2028. Further, over the past two years, according to a Bloomberg tally, Amazon has committed to spending $148 billion to build and operate data centers around the world. The company plans to expand existing server farm hubs in northern Virginia and Oregon as well as push into new precincts, including Mississippi, Saudi Arabia and Malaysia.

Bloomberg reported that Blackstone Is Building a $25 Billion Empire of Power-Hungry Data Centers. According to the article, “the private equity giant says landlord QTS could be one of its best investments ever”.

This is a global phenomenon. According to Reuters, “CBRE forecasts average vacancy rates across Europe's biggest markets - Frankfurt, London, Amsterdam, Paris and Dublin - will drop to a new low of 8.2% in 2024 after ending 2023 at a previous all-time low of 10.6%.” Secondary markets are not faring much better.

Meanwhile in Asia, according to Japan Times, “demand in Southeast Asia and North Asia is expected to expand about 25% a year through 2028, according to Cushman & Wakefield data. That compares with 14% a year in the U.S.”

Prediction 3 - Consequently, energy use will also see incremental demand on top of the previously forecasted demand for hyperscalers and EVs. This will further push demand for sustainable energy generation, efficiency in generation using traditional sources, efficiency in distribution and electrification use cases at point of use. Regulation will ramp up.

Q2 Outlook - Done deal. This has happened and will gain pace.

This is probably the most meaningful of the predictions. At the current trajectory, energy generation, storage and consumption will all undergo rapid change driven by AI, As the value cases from AI growth fund investments, research and growth; societies all over the world will benefit from second order effects. I write about this topic extensively in essay # 13, “Energizing AI: Navigating the Power Dynamics of Tomorrow's Tech” but new evidence keeps appearing supporting my claim. Take for example, Equinix’s letter of intent to purchase power from Oklo’s planned ‘powerhouses’ to serve Equinix’s data centers in the US on a 20-year timeline – at rates decided in future Power Purchase Agreements (PPAs). Equinix will have the right to renew and extend PPAs for additional 20-year terms. Meanwhile the current draw on electricity continues unabated, last year Blockchain firm Standard Power announced plans to procure 24 SMRs from NuScale for two US data center sites.

As I have previously written, Microsoft has previously signed a nuclear carbon credits deal with Ontario Power Generation for its operations in Canada, now we read that they have recently signed an energy agreement deal with nuclear fusion startup Helion. I also wrote previously that they had signed a 24/7 nuclear power deal with Constellation to power its Boydton data center in Virginia.

Google on the other hand is not much behind, an example is its use of a new Sonoran Solar Energy Center, developed by Salt River Project and NextEra Energy Resources in Arizona, which pairs a massive 260-megawatt solar farm with an even more impressive one gigawatt-hour battery. This center, combined with another 88-megawatt solar-and-storage facility in Arizona, could power 80,000 homes and store enough for up to four hours.

The US government seems to be paying attention. Energy Secretary Jennifer Granholm told Axios that the growing demand for power from AI and data centers is a "problem" that needs to be addressed, while presenting a loan guarantee to help restart a shuttered nuclear power plant on the shores of Lake Michigan. If this is any indication, nuclear generation could start becoming more than the 20% it is currently at for US power generation.

Prediction 4 - Talent fragmentation will exacerbate - Individualism especially derived from professional experiences of the last three years will resist corporate apathy and back to office regimes.
Specifically, the legion of pandemic era graduates will push against accepted work norms from the “before times” leading to a different trajectory in workplace norms, the orthodoxy will continue to push back.
On top of that, hiring excesses during the pandemic rebound and the layoffs recently eroded trust in corporate regimes.
Due to this struggle, talent will fragment between the two extremes, the most capable folks will be drawn to a more effective middle path which will become clear in 2024 and as is typical, will most likely be pioneered by smaller companies and startups.
In addition the most entrepreneurial folks and those with access to capital will further remove themselves from the talent pool available to corporations because of the “Silver Tsunami”. The silver tsunami metaphor describes the aging population, particularly baby boomers reaching retirement age. In this wealth transfer, people will increasingly choose small and medium enterprises or entrepreneurship.

Q1 outlook - Unclear, we will need to wait and see.

So far the trend of layoffs and force reduction barring the AI space has not let up. We will have to wait and see what unemployment (which improved in the most recent report), inflation and rate cuts do to the economy throughout the year. Then there are the elections.

Prediction 5 - Democratization of capital will accelerate. In 2024, the industry and platforms in support of equity crowdfunding catalyzed by the JOBS act of 2012 will hit prime time. The top income + Net-worth decile (or more?) Americans will qualify as “accredited investors” and will deploy even more capital through these platforms taking liquidity away from traditional channels. Combined with the prediction 4 above, this will become a powerful force in 2024.

Q1 outlook - Unclear, we will need to wait and see.

In addition to many “promoters” syndicating capital from accredited investors who buy and roll up investments in asset based projects, the signs that this might bear out to be true is that a secondary market has emerged. According to WSJ, large firms like KKR, Ares, Apollo and Blackstone et al are pushing aggressively into “asset-based finance,” a kitchen sink of debt including loans backed by equipment such as fiber-optic networks, data centers, storage units, warehouses etc.

Prediction 6 - Even though talent will fragment, unit productivity will significantly improve in jobs requiring knowledge work through extensive AI enabled support tools. Technology platforms will take the lead in 2024 especially @Salesforce, @Microsoft, and possibly also @IBM among others. They will use the old technique of bundling and offering these “assistants” or “intern-bots” as features. Meanwhile, small and medium-sized businesses will choose newer, more affordable unbundled technology stacks with significant impact.
The nature of the technology market will start fundamentally changing with the center of gravity moving to the post Cloud, Mobile, Social Media era. We will have new leaders, new economics and possibly different distribution ecosystems. We should see the new landscape clarified by the end of the year.

Q1 Outlook - Trending as predicted.

It's amazing how quickly this adaptation has been incorporated into business practices not just in the US but across the world. A casual browse across Youtube or other short form video clearly shows the widespread use of AI generated imagery and videos. It's apparent but hard to quantify.

Interestingly, recently (late 2023) results of a natural experiment in italy were published. It all started when in early 2023, Italy’s Data Protection Agency took the extraordinary step of banning ChatGPT within the country over concerns about consent and personal data privacy. David Kreitmeir and Paul Raschky at Monash University in Australia downloaded data from GitHub showing the hourly coding output of 8000 programmers in Italy and in two other European countries from before and after the ban. The results are something of a surprise. “We find that the ban of ChatGPT decreased output of Italian GitHub users by around 50% in the first two business days after the initiation of the ban,” say Kreitmeir and Raschky. “Output levels returned to normal levels after that.” The researchers say the return to normal was probably the result of coders finding ways round the ban, which is easy to circumvent using a virtual private network or Tor Relay to access ChatGPT in other countries.

In february, Klarna announced that its AI assistant powered by OpenAI had 2.3 million conversations, two-thirds of Klarna’s customer service chats in one month, or the equivalent work of 700 full-time agents on par with human agents in regard to customer satisfaction score. According to the press release, it is more accurate in errand resolution, leading to a 25% drop in repeat inquiries.

In their earnings call, Palo Alto Networks mentioned roughly halving the cost of their T&E servicing, and ServiceNow mentioned increasing our developer innovation speed by 52%, clear data are emerging which point to the prediction being right but it's early. It will be interesting to watch this space through the year.

Prediction 7 - The traditional stacks that underpinned the digital ecosystem since the start of this century start unraveling. Interest and intent based audiences will get tainted by GenAI this year. The content produced, how it is consumed and what a user is, will all become suspect before the end of the year with plainly evident outcomes. According to @The Economist, in 2024 countries with more than half the world’s population - over four billion people - will send citizens to the polls. Everything that AI can do to achieve this audience tainting and content manipulation to influence opinion will hit primetime in these elections. This is a rare confluence of events and this technology’s potential will be too irresistible for it not to be exploited.

Q1 Outlook - Trending as predicted, unfortunately.

In January, I was already seeing signs of this trend in many elections from US primary elections to the national elections in Pakistan as I mentioned in this LinkedIN post. Elections in important democracies such as the US, India and the UK are later this year but early signs are merging of more malfeasance afoot. An Al Jazeera review and subsequent forensic testing found at least three instances of AI-created or altered content published on the official Instagram handles of both the Indian National Congress and the BJP, two major political parties in India, since February. Political parties are pushing the limits of AI’s use to both ridicule opponents and boost their own popularity on official pages, and in doing so, testing the boundaries of platform policies on labeling deceptive political content. Expect more reports of this through the year.

In Indonesia, the world's third-largest democracy, elections were conducted in February and offered a preview of how generative AI can be used in political campaigning. There wasn't large-scale malfeasance but yet the use of AI was reportedly widespread and the boundaries and term sof use were tested in many ways. The winner and president-elect Gen. Prabowo Subianto - a once-feared former special forces commander, reinvented himself as a cute and cuddly character using AI. Posts tagged with his hashtag got more than 19 billion views on TikTok. About half of Indonesia's 205 million voters are under 40.

Prabowo’s AI likenesses were reportedly created using tools from Midjourney Inc. Other uses included the creation of the Pemilu.AI app which reportedly uses OpenAI's GPT-4 and 3.5 software to craft hyper-local campaign strategies and speeches.

It wasn't all clean. For example, "deepfake" videos of the late strongman ruler Suharto urging voters to support its candidates who all belonged to the Prabovo camp. The clips containing campaign messages were marked as generated by AI. It is reported that Suharto's image was created with Midjourney and imaging app Leonardo AI, while his voice was crafted using U.S. voice cloning startup ElevenLabs's software combined with in-house technology.

Just this month, Microsoft said it detected a surge of more-sophisticated AI tools in the January presidential election in Taiwan, including an AI-created fake audio clip of a former presidential candidate endorsing one of the remaining candidates. That marked the first time the technology giant’s researchers on threats had seen a nation-state actor using AI to attempt to influence a foreign election.

As I said in that post, “it's going to be a rough year”.

Prediction 8 - Governments will get increasingly involved in ensuring that they do not miss the AI boom while at the same time ensuring that it doesn't negatively affect sovereignty or governance. This will be a very passive aggressive year for policy, regulation and governance; countries and various nation blocs will appear to be cooperating while they simultaneously push parochial advantage. This will negatively affect free markets globally. In combination with Prediction 7, this will be a potent mix.

Q1 - Trending as predicted, mostly correct.

Last year I wrote extensively about this trend in essay #11, “AI's Rulebook Emerges: Deciphering Regulatory Realms and Policy Pioneers”. Since then, the Artificial Intelligence Act that I had discussed was in fact adopted by the EU Parliament on March 13 and is expected to soon become law when it is passed by the European Council. It will take up to 24 months for all of it to be enforced, but enforcement of certain aspects, such as the newly banned practices, could start to happen in as little as six months. From my reading, it's going to feel a lot like GDPR.

Meanwhile governmental intervention is not just on the side of control. MacroPolo, a think tank located at the Paulson Institute in Chicago, USA, reported that almost half of the top researchers in AI research, who are in the top 20% of the world, are from China. With encouragement from the Communist government and a dedicated focus to lead in this arena, China has added AI programs in more than 2,000 faculties, more than 300 of which are at top universities, according to Damian Ma, managing director of MacroPolo. Further as previously mentioned, the Chinese government has made AI a national priority through various policies, funding initiatives and strategic plans. Notably, the "New Generation Artificial Intelligence Development Plan'' unveiled in 2017 outlines China's vision for becoming a global AI leader by 2030. As part of this plan, AI models have to be pre-registered with the government but need to comply with “socialist principles''. This compliance is rationally impossible making it an instrument of control where the government can manage the haves and the have-nots at its discretion. I won't even write about the chip embargo from the US to China.

Let's see what the rest of the year brings.

Prediction 9 - AI related litigation will start establishing legal precedent forming legal guardrails around AI use from intellectual property to privacy. By the end of 2024, there will be established case law and rulings that will start forming the regulatory bedrock for the evolution of this industry.

Q1 - Trending as predicted, mostly correct.

There isn't a whole lot more I can say, given my recent essay #16, “Judging a New Technology”. I am watching this space, you should too.

—----

Sources and References:

The Sophophile Writes

Discussion about this post