Senior Industry Executive: This wave of AI is just beginning, it's far from reaching its peak.

After the explosion, can AI create brilliance again?

Recently, Elad Gil, known as a Silicon Valley startup mentor, published a blog post titled "The Early AI Era (and the AI Hype Cycle)". The article points out that we are still in the early stages of AI, and the peak of hype and impact is yet to come.

Gil states that AI is a step function for the introduction of new features and products, marking the arrival of a new technological era, rather than a continuation of CNN/RNN/GAN. It's like when cars already existed, someone invented airplanes and said, "Airplanes are just cars with wings," without mentioning all the new use cases and their impact on travel, logistics, defense, and other fields. The era of aviation is about to begin, not just a "faster car era."

Gil believes that given the planning cycle of large companies usually takes 3-6 months, and the prototyping and construction of large companies take a year, we are still far from the peak of AI usage or the peak of AI hype.

The following section is translated from the blog post:

I have worked on the development of early machine learning (ML) systems and products at Google and Twitter (after Twitter acquired Mixer Labs, a location-based technology developer co-founded by me). In the following decade, I have been a founder and executive of machine learning companies and made investments.

Before the rise of new AI architectures, especially those based on Transformer and diffusion models, almost all machine learning startups failed. In the previous AI wave, the value mainly flowed to established companies rather than startups because their capabilities were not advanced enough to create new market opportunities.

This is a PPT slide I used around 2017-2019 (borrowed from Brandon Ballinger), which reflects the world of CNN/RNN/GAN in the previous machine learning wave.

Today, when many business people talk about "Artificial Intelligence (AI)," they see it as a continuation of CNN/RNN/GAN. In fact, it is a step function for the introduction of new features and products, marking the arrival of a new technological era.

It's like when cars already existed, someone invented airplanes and said, "Airplanes are just cars with wings," without mentioning all the new use cases and their impact on travel, logistics, defense, and other fields. The era of aviation is about to begin, not just a "faster car era." (Of course, we should fully recognize the importance of the previous machine learning and deep learning waves to all of this - however, considering it as a continuous extension may overlook the qualitative change of this technology).

This is the PPT I'm currently using.

The launch of GPT-3 in June 2020 marked the occurrence of some interesting things. GPT-3 is a huge improvement over GPT-2 and previous models. It is not yet capable of accomplishing everything we now consider as "AI," but it strongly suggests what is about to happen (a few months later, I discussed GPT-3 on the A16Z podcast because it was so remarkable).

For those who are aware, the release of GPT-3.5 in March 2022 solidified the belief that Transformer-based models are the future trend. Within companies like Google, OpenAI, Microsoft, and Anthropic, early exposure to the model provided some individuals with a preliminary understanding of what is to come. An engineer at Google eventually announced an internally developed AI chatbot called LaMDA, which is "perceptive" - this chatbot is a precursor to products like chatGPT and Character.AI.

The real starting gun for this AI wave was propelled by two sets of releases. First, the launch of image generation products like Midjourney and Stable Diffusion, followed by the shocking release of ChatGPT a few months later, which sparked the public's imagination and became a moment of explosive growth for AI startups. ChatGPT truly exemplified the capabilities of these new types of AI and the power of RLHF (reinforcement learning from human feedback) in language models.

The launch of ChatGPT made a significant impact, making people realize the profound significance of AI in new functional areas and igniting the popularity of generative AI. This release happened 8-9 months ago, while GPT-4 was introduced only 5 months ago. Considering that planning cycles for large companies typically take 3-6 months, and the prototyping and construction of large companies take a year, we are still far from the peak of AI adoption or the peak of AI hype. Most large companies are still trying to understand what "artificial intelligence" means for them, and there is still a long way to go before embracing this new technology.

True enterprise adoption will still take several quarters or years. In fact, there are likely at least 4 waves of artificial intelligence to consider.

Wave 1: Native GenAI companies. ChatGPT, Midjourney, Character.AI, Stable Diffusion, Github Copilot, and other early products have now gained significant revenue and user traction. Clearly, there are some great machine learning companies that predate GenAI and continue to participate in the current era (Hugging Face, Runway, Scale, and WandB are a few that come to mind).

Wave 2 (current stage): Early-stage startups and fast-growing mid-market existing companies. This is the first wave of startups built on top of GPT-3.5/4, such as Perplexity, Langchain, Harvey, and others. At the same time, a few founder-led multi-billion-dollar companies like Navan, Notion, Quora, Replit, and Zapier quickly launched AI-driven products, becoming early adopters of this wave. Microsoft, Adobe, and Google are notable outliers as large enterprises are rapidly shifting towards AI—Microsoft possibly due to its internal ties with OpenAI, and Adobe because diffusion models tend to be cheaper and easier to train than large language models.

Wave 3 (upcoming): The next wave of emerging startups. This will be exciting as it may include new formats like voice and video, as well as new infrastructures, in addition to using natural language in more ways. Companies like Eleven Labs/LMNT/LFG Labs, Braintrust, and others will offer progressive experiences. A large wave of new startups is on the horizon, with just the current YC investment batch seemingly having 100 or more.

Wave 4 (coming in 2024/2025?): The first wave of large enterprise adopters. Due to long planning and build cycles, it is expected that the first real products (as opposed to PPT presentations or prototypes) will emerge in one to two years from companies other than Microsoft, Adobe, Google, and Meta. By then, the revenue of AI infrastructure companies will start to grow significantly, hype will reach its peak, and we will see further acceleration in AI investments.

The future is bright.

This new wave of technology has enormous potential impact on humanity. For example, Google's MedPaLM2 model performs significantly better than human doctors to the extent that it makes medical experts worried about the model worsening the situation!

Given the immense potential of this technology, it will be exciting to one day witness the breakthroughs it brings in education, healthcare, enterprise and consumer software, and other aspects of life, leading to significant innovation. Now, it's only been 8-9 months since ChatGPT awakened the world and ushered in a new era of AI. As we continue on this nonlinear technological journey, exciting times lie ahead. We are still in the early stages of artificial intelligence, and the peak of hype and influence is yet to come. There is much more to be done.

Elad Gil: Prominent Angel Investor and Serial Entrepreneur

Gil is a renowned angel investor and serial entrepreneur. He joined Google when it had just over a thousand employees and left when it had grown to over fifteen thousand. During his four-year tenure, he created Google's mobile team, served as the first product manager for important products, and was involved in all aspects of team building and three significant acquisitions.

Gil also helped Twitter grow from around 90 employees to a industry giant with 2500 employees. As the VP of Strategy at Twitter, he played a key role in the company's scaling efforts, including product development, platform expansion, internationalization, user growth, mergers and acquisitions, recruitment, and organizational processes. Even after leaving, he continued to advise Twitter as a CEO and COO, providing valuable business insights.

In the investment field, Elad has a keen eye for opportunities. Around 25 of the companies he has invested in have become "unicorns." Moreover, the companies he has been involved in, such as Airbnb, Wish, Coinbase, PagerDuty, Square, and Pinterest, have all successfully gone public.