Crucial Intel for Entrepreneurs: The Top 3 GenAI Predictions You Must Know!
(1) Brace for AI Regulations and Consequences, (2) Embedded AI is the Way to Grow, and (3) Massive Content Dilution is Imminent
Generative AI (GenAI) has taken the world by storm, and for good reason — it has the potential to permanently change the way most people work and also create significant value for the global economy. A study examined 2,100 tasks across 850 jobs around the globe and predicted that current generative AI and other technologies have the potential to automate work activities that absorb 60 to 70% of employees’ time! The extent of the disruption for the average knowledge worker is reminiscent of the impacts of the iPhone and Windows! Whether this technology will create or destroy jobs is up for debate, but what is clear to me is that most businesses will leverage GenAI in some dimension.
However, keep in mind that GenAI is still a small part of the broader AI pie! It is estimated that GenAI only accounts for ~16% of the total economic potential of AI. GenAI is still a subset of a subset of AI. And of course, it is an industry that has received an overwhelming amount of attention as of late. But this isn’t to suggest that GenAI will fade into irrelevance. In fact, it’s hard not to get excited about GenAI! As a founder or entrepreneur, imagine a future where your company’s:
1. entire customer service operation is handled by LLMs
2. marketers use image generators (such as DALL-E) to create infinite hyper personalized ads
3. software engineers use LLMs to generate code thus becoming 30–50% more efficient
4. scientists in R&D use GenAI tools to generate more drafts and designs cutting 10–15% of R&D costs
I specifically chose those four functional areas as examples since they are expected to account for 75% of the potential economic value from GenAI. Given the stunning possibilities with this technology, it is no wonder founders and entrepreneurs are either busy building GenAI or incorporating it in their own businesses! Earlier this year I laid out 5 general predictions on AI during Collision week. Now, I think it would be helpful to tailor some GenAI considerations specifically for builders who are creating GenAI applications or incorporating it into their current product offering!
(1) Heads Up! Build Responsibly or Brace for Incoming Regulations
Most entrepreneurs building GenAI or incorporating it into their business are obviously aware that just like any other technology, GenAI is subject to existing measures of privacy, copyright, and accountability. However, unlike other technological advancements to date, AI poses significantly higher risks! Hundreds of scientists, tech industry execs and public figures signed an open letter warning that AI could lead to an extinction level event. Notable signatories includes Sam Altman (CEO of OpenAI), Geoffrey Hinton (Godfather of AI), Jaan Tallinn (Skype co-founder) and Max Tegmark (author of Life 3.0). Founders and entrepreneurs need to anticipate regulatory firewalls to future proof their innovations.
Social acceptance from the public is one of the main reasons that AI is proliferating and applications like GPT were able reach record users in a short time. People have already been acclimatized to AI in everyday life through features such as Google autocomplete or Tiktok feed recommendations! Incoming regulation around AI aims to codify the social contract between builders and users. The EU, through the “EU AI Act”, is currently leading the charge around comprehensive laws and legal frameworks which will set or highly influence the global standard through the “Brussel Effect”.
“With this Act, the EU is taking the lead in attempting to make AI systems fit for the future we as human want,” — Kay Firth-Butterfield, the Head of AI at the World Economic Forum.
The act calls out GenAI specifically and highlights concerns around transparency. The act A) requires disclosure of content that was generated by AI (like food labeling standards), B) prohibits illegal content generation, and C) requires detailed summaries of copyrighted training data. The final draft of the act will likely come into effect in 2–3 years (at the time of the writing), but the EU and the US are working together to draft an “AI Code of Conduct” in advance of formal regulations to guide responsible AI development that will come in weeks or months rather than years.
These regulations are of the utmost importance to founders as violating them could lead to a fine of €40MM or 7% of revenues (which ever is higher) for prohibited AI systems. A lower tier fine (€20MM or 4% of revenues) would apply for transparency violations. This fine is remarkably higher than GDPR violations which is €10MM or 2% of revenues and should be taken seriously.
(2) Supercharge Growth and Retention Through Embedded AI
If you are a PM or manager, you’ve probably already asked yourself how you can incorporate AI into your product. Most application layer products around GenAI are built to improve existing workflows rather than replace workflows entirely. For example, LLMs are built to augment SWEs in writing code, rather than replacing whole technical teams — at least for now! Most applications are currently built this way as embedding GenAI features into existing workflows will optimize for A) user adoption, B) data network effects, and C) increased stickiness.
A) Embedded GenAI is an effective way to accelerate user adoption as the product can seamlessly integrate into existing workflows. There are already several examples of successful embedded GenAI applications (think co-pilot-as-a-service). Consider Adobe Photoshop and Stable Diffusion for embedded image generation, Snapchat and MyAI for embedded LLMs or Github and Copilot for embedded code generation. Added to the point, many GenAI companies are currently building point-solutions that are difficult to scale standalone without embedding it in another workflow. Even freestanding GenAI products often require several integrations or APIs to become sticky. Embedding your GenAI product into other softwares or existing workflows will give the product a head start.
B) Data networks effects and user adoption goes hand in hand. More users, means more proprietary data, and more user feedback through reinforcement learning. It is a positive feed back loop that ultimately increases accuracy and makes higher quality predictions. Consider the virtuous loop that ChatGPT created by releasing a free version to the public that constantly collects user feedback. In addition, hallucinations in GenAI are routine, and accuracy has now become a requirement. For instance, a CEO of an AI company told me recently that some government customers will only use AI software if they can reach “five nines” accuracy (aka 99.999%). Embedding a product to develop data network effects could build an enduring moat.
C) Embedded GenAI tools provide a better value proposition to users which increases stickiness. Customers benefit from a contextual and seamless experience. The most notable example is the Microsoft Office Suite and Copilot (to be released). Copilot will be able to leverage existing information from Outlook, Powerpoint, Excel, Teams, and more without interrupting normal course business functions. Like the embedded finance storyline, users don’t want to migrate from their current trusted platform due to risk. Embedded GenAI features add another compelling dimension without interrupting user experience.
(3) Survive the Content Flood: Massive Dilution is Imminent!
Sales and marketing is one the biggest use cases for GenAI and is expected to create ~$949Bn of economic value and research suggests that ~20% of current sales-team functions could be automated! Imagine a future where marketers will be able to leverage tools for:
- Lead Generation: identifying leads or enriching leads
- Marketing Optimization: A/B testing, SEO, website structure/navigation
- Hyper Personalized & Automatic Outreach: emails, chats, social media
- Creative Content Generation: blog posts, social media posts, Medium articles (note: this article was written without AI generated content)
In my opinion, the huge gains in productivity come at the cost of content dilution. Anyone who has played around with GPT or Jasper.ai will be amused by its ability to produce high quality marketing content at minimal cost. However, the outputs of mass generated GPT content seems stale, uninteresting, and formulaic. The “economics of abundance” POV argues that our present economic system only finds value in scarcity. This theory explains the phenomenon in the booming video games industry around skins and in-game cosmetics. For decades, gaming was built around the one-time purchase of a game that you played for years. However, free games such as Fortnite, allowed game developers to quickly reach a wider audience. The video game skins industry converted a scarce resource (one-time purchase of games) into an abundant resource (free to play), which then increased demand for a related scarce resource, in-game cosmetics. The advent of video game skins and other in-game cosmetics created a $40Bn industry!
Several sources already refer to mass generated AI content as “shallow”. In addition, we are also inundated with information overflow via email and social media. Now imagine your entire feed filled by automatically generated Tweets, or posts devoid of originality…! Similar to the in-game cosmetics example, the good news is that abundance of generated content also creates a new relative scarcity. If zero marginal cost marketing content becomes abundant, then I predict that it will create a newly scarce resources in the form of: original content that is certified authentic (i.e. GenAI free) and grounded in creativity, real life relationships, experiences. A federal judge in the US also agrees. Last month Judge Beryl A. Howell ruled that GenAI does not have copyright protection as it “lacked human creativity” - which sets an important legal precedent. As long as you are still selling to a human, I urge you to tread carefully when using shallow generated content that comes at zero marginal cost.
AI is Going to Eat Software
Marc Andreesen famously said that “software is eating the world” in 2011, and ushered in a golden decade for software. The software industry experienced a transition from analog to digital processes, gathered vast amounts of data for insightful analytics, and rewarded disruptive entrepreneurs with soaring valuations. For more on this see Robert Smith’s podcast episode on software. In counter to Andreesen, Jensen Huang (CEO of NVIDIA) retorted, “Software is eating the world, but AI is going to eat software”. For anyone building GenAI products, or incorporating it into your existing products — I applaud you. I believe GenAI and AI will be a category defining technology, and will inspire significant advancements in productivity and transform the nature of work.
Let me know if you agree or disagree with my views! I’m eager to engage in thoughtful dialogue, and connect with leaders in this space.
The inevitable pinnacle for all tech lies in the beginning: information technology, itself.
AI is simply the evolution of the core process of information, and the more we discover the mechanics of Large Language Models and how it processes information, the more we will come to grasp how human beings have been pre-trained to process information. The issue with every pre-training model and computational algorithms is one, elementary; the inherent bias within its subjective raw data.
The value of AI will not be found within external applications, entertainment or any products but within the very thought process we apply with every impulse over the course of our lives to interpret ourselves and the world around us, which evolves with the growing access of information and cascading technical tools to process that information. The key is our ability to process information, not the computers or any and all consumer products.
This shift will reveal the deepest conceptual blindspots, underlying patterns and flaws woven and cemented within the foundations of all human nature, and within our subconscious preconceptions. Over time, it will reflect a greater sense of cognitive pattern recognition that we have yet to achieve while empowering every mind to process the manipulation tactics and methods that have been employed to take advantage of each and every organic processor, by reducing each of our capacity to challenge or even perceive greater contexts than what has been governed by traditional common sense and government/financial systems.
AI will give every human being moving forward the ability to open their mind, process information objectively from their point of view and evolve their perspective of information. A realization of the perpetual affects of generational subjectively-governed thought processes and frames of reference.
Exciting times for the open minds, ready to process information and their experiences from an entirely new and healthier vantage point than: Think this way or else.