Goldman Sachs In-Depth Report: Is generative AI hype or real change?

Author: Bu Shuqing

To say that this year's rapid growth of the sector, AI must rank first.

Driven by the wave of investment in AI technology, the market value of the "Seven Giants" of Apple, Microsoft, Google, Amazon, Meta, Nvidia and Tesla has soared by 60% this year to an astonishing $11 trillion.

And this "rocket" jump has driven the Nasdaq to surge 34% this year, dominating most of the S&P 500's gains this year. **

** **The concept of AI has risen to the sky, and doubts have begun to appear frequently. The loudest of these is, is the AI hype too much? Does Generative AI Really Have Disruptive Potential? Is it worth the huge investor enthusiasm right now?

In the "Top of Mind" report recently released by Goldman Sachs, Goldman Sachs strategists Allison Nathan and Jenny Grimberg conducted in-depth dialogues with a number of AI industry professionals, including Sarah Guo, founder of AI venture capital firm Conviction, New York University professors, start-up companies Robust.AI CEO and founder Gary Marcus, Goldman Sachs software and Internet analysts Kash Rangan and Eric Sheridan, tried to answer the above questions.

In addition, they also discussed the most attractive investment opportunities in the AI field and the risks that investors should pay most attention to.

Revolutionary Changes in AI

The fundamental difference between generative AI and traditional AI technology is that the former creates content by understanding natural language, while the latter relies on programming languages. According to Goldman Sachs software analyst Kash Rangan,** this is the key transformative feature of generative AI technology . **

First, it is capable of generating new content in the form of text, images, video, audio, and code, whereas traditional AI systems train computers to make predictions about human behavior, business outcomes, and more. Second, it allows humans to communicate with computers in their own natural language, which has never been done before; traditionally, computers use programming language prompts.

Guo further explained that in the era of software 1.0, humans needed to write code to perform specific tasks, and in the era of software 2.0, they trained neural networks by collecting data "hardly". Now humans ushered in the era of software 3.0:

The underlying model is available via open source or API, with natural language capabilities, reasoning capabilities, and possesses common sense about the world. In this model, companies don’t need to collect nearly as much training data, making technology more useful, more accessible, and less expensive.

Since ChatGPT exploded out of the circle last year, many people have felt the power of generative AI technology. Analysts believe that generative AI may reshape the way social production works and add a new growth engine to the global economy.

According to Guo, the transformative potential of generative AI has already begun to be translated into reality. Any AI investing firm can now invest in these models, enhancing their business or transforming it.

Rangan estimates that, in some cases, top developers have seen productivity gains of 15-20% through the use of generative AI tools. **

With the popularization of AI, Guo predicts that in the future, more fields, especially traditional service industries such as law, data analysis, picture, voice and video generation, will be increasingly served by AI.

Goldman Sachs TMT industry analyst Peter Callahan pointed out that retail investors believe that generative AI technology has all the elements of platform transformation, and may change the experience of enterprises and consumers in all aspects.

Separately, Joseph Briggs, senior global economist at Goldman Sachs, said this transformative potential could have profound macroeconomic implications.

He estimates that the popularization of generative AI technology in the United States and other advanced economies in the world can increase the annual labor productivity growth rate by about 1.5 percentage points in the next 10 years, and the global GDP will eventually increase by 7%. **

Goldman Sachs U.S. equity strategists Ryan Hammond and David Kostin believe that U.S. stocks will also benefit from this, and a broader rebound is expected in the medium to long term. ** The fair value of the S & P 500 index will be about 9% higher than it is now. **

** **

Artificial intelligence is far from intelligent enough, be wary of excessive hype

In the long run, the transformative nature of AI technology is beyond doubt, but given the current development progress of this technology, is the market hype about it too much?

Marcus's answer is "yes" because **"Current artificial intelligence is far from intelligent enough". **

He pointed out that the so-called neural networks of current AI function completely differently from the neural networks of the human brain.

Although AI can perform "reflexive" statistical analysis, it has almost no mature reasoning ability. These machines can learn, but much of it revolves around the statistics of words and correct responses to cues, rather than abstract concepts. And, **they don't have an "internal model" that allows them to understand the world around them like humans do. **

Marcus issued a warning to investors:

Be wary that AI performance is not as amazing as many people think. I wouldn't say it's too early to invest in AI; some investments in companies with smart founding teams and a good understanding of product-market fit may succeed, but there will also be plenty of losers.

Artificial general intelligence (AGI) may eventually be possible,** but humans are nowhere near that goal, and no amount of investment is likely to change that, Marcus said. **

In addition, investors can also learn some lessons from history.

Goldman Sachs market strategists Dominic Wilson and Vickie Chang have mentioned that during past innovation-led productivity booms, for example, after the spread of electricity (1919-1929), personal computers and the Internet (1996-2005), stock prices and valuations soared to form bubbles , and eventually breaks down.

Guo argues that, even today, some areas of the private equity market are still mispriced. While investors have a greater understanding of these areas, the same approach to investing is still generally employed.

Misjudging the timing of change is a common pitfall in investing, she warns. **As an early stage investor, she is less concerned with valuations and instead picks markets, products and businesses that she believes make sense.

Goldman Sachs internet analyst Eric Sherida takes a slightly different view.

He believes that the vast majority of outstanding AI stocks are still trading at reasonable multiples of GAAP EPS.

Rangan also believes that AI may not be in the hype cycle, because this wave is dominated by technology giants, not start-ups:

This technology cycle is not dominated by (AI) upstarts and is unlikely to end anticlimactically or take a long time to get started. The transition from mainframes to distributed systems (computers) in the early 1990s, and from distributed to cloud computing in the early 2000s, took longer than many expected, both because large, established companies were key voices of opposition.

As Rangan said, OpenAI, the company behind ChatGPT abroad, is supported by Microsoft, Google launched Bard, invested in AI start-ups such as Anthropic, Meta launched LLaMA, domestic giants such as Baidu and Ali have also released their own models, and the global AI competition is in full swing. middle.

"Pick and Shovel"

There are endless voices questioning the hype. What are the most eye-catching investment opportunities in AI at present?

According to Rangan and Sheridan, the opportunity lies not just with the big tech companies that develop the underlying AI models,** but with “Picks and Shovels” businesses. **

"The pick and shovel" is one of investing legend Peter Lynch's preferred investing strategies, investing in companies that indirectly benefit from a boom.

Rangan and Sheridan argue that companies that serve the field, such as semiconductor companies, cloud computing hyperscalers, and infrastructure companies, can all be well-positioned in the current "build" phase of the current AI boom.

Guo feels similarly, but also sees opportunities across the stack,** and is most excited about the application layer. **

Many investors are uncertain about this layer, thinking that all the value is in the model training itself, but it will take a lot of creativity and work to get non-deterministic models to work in production use cases. There are many areas in which startups and existing app companies alike will take advantage of these capabilities...we're excited.