From Hype to Business: Key Elements in Unlocking Generative AI's Value for Business

Since their launch, Generative AI platforms like OpenAI's ChatGPT, Google's Bard, and others have generated (no pun intended) massive hype. Beyond the hype, we need to sober up and ask the quintessential questions; what will be the impact of Generative AI on business, now and in the future? How can the (Generative) AI industry unlock value for the digital economy?

Competitiveness in the digital era has often been underlined by how quickly companies develop (for digital natives*), or adopt (for non-natives) ground-breaking digital technologies to gain a market advantage. In the case of OpenAI itself, launching ChatGPT on 30 November last year was its first-mover advantage that resulted in the platform achieving 100 million active monthly users in just two months. For context, it's worth noting that it took TikTok nine months to gain this many users, and over four years for Facebook. This earned ChatGPT the unequivocal title of "the fastest-growing consumer application in history " and with it too a valuation of US$29 billion.

This first-mover advantage was true for OpenAI as a native Generative AI technology company launching ChatGPT, but it can also be true for companies that move quickly to integrate this technology into their business operations. In the automotive industry, the commercialization of electric vehicle technology in 2008 created massive outperformance for Tesla gaining more than a 7x advantage in stock value (or peak market capitalization of more than US$1.2 trillion in January 2022) ahead of its gasoline-powered counterparts. The swift adoption of this technology (and business model) by fast-follows like Li Auto, Rivian, and others has created a US$193.6 billion industry in which these players enjoy healthy returns and guarantee future sustainability. This was true for Netflix adopting online streaming technology in 2007 and Dominos Pizza implementing innovative digital engagement channels. Whether it is developing or integrating new technology, it's the first movers (with the correct elements in execution of course) who gain the biggest share of the cake.

Should companies be bothered by Generative AI?

The age of Generative AI is here!

The rate of user adoption, level of investment, number of release cycles, and the preliminary SaaS use cases in Generative AI have heralded a new technological paradigm in just the last six months. All these are important indicators of how seriously this technology should be considered by all stakeholders; from native AI companies, users, governments, right to the corporate world.

Unlike the typically complex and innate use cases of AI; in healthcare with algorithmic diagnostics, customer service virtual assistants, self-driving cars, or voice recognition, Generative AI has truly democratized AI. Virtually anyone with an internet connection and access to a Generative AI platform can input a prompt and generate new content. This is made possible by the cutting-edge Large Language Models (LLMs) (artificial neural networks with billions of parameters) that power the technology. The LLMs can produce different types of content including text, images, audio, and video in response to user prompts. Unlike 'traditional' AI models that are trained on localized, structured, and labeled data, LLMs are trained on the vast public data that sits on the internet. This makes Generative AI technology revolutionary. We have had the Internet of Everything (IoE); Generative AI is the AI of Everything (AIoE). Anyone can develop new content, in any format, that has never been seen anywhere, using a straightforward consumer interface in just a few seconds!

Software engineers, for example, can input a plain text code request in ChatGPT and receive whole blocks of functional code. As you can imagine this increases the efficiency of writing code, reduces errors, and improves the shipping time for bug fixes and new product features. Another scenario is for product managers who may be interested in synthesizing thousands of user reviews into pipeline features that truly address user problems. Product managers may input prompts for this task in a Generative AI platform. As long as the model data is available, the platform will provide suggestions for improving the product. This greatly enhances the PM's productivity which otherwise could take weeks or months in data analysis, brainstorming sessions, and product experimentations before arriving at a final iteration scope. The genius of Generative AI platforms is that any professional; a lawyer, a medical researcher, a journalist - virtually anyone in any use case, can use this technology to improve their work.

In my last article, I argued how flaws in content moderation on online social platforms like Facebook, Twitter, and TikTok represent the biggest modern crisis in big tech . I highlighted how more expansive algorithmic moderation is a sustainable solution for the crisis. I have no doubt that integrating Generative AI in these platforms, with multimodal capabilities in both its comprehension, and development of natural language, images, audio, or video can significantly improve content censorship and create healthier online communities. I imagine that this is a different, but related dimension to why Microsoft integrated ChatGPT within Teams - to synthesize content shared on the platform.

Admittedly, it's early days for Generative AI, and the technology has struggled with bias and accuracy in some cases. But the potential for the technology to be at the center of building new technology, improving existing products including in the pharmaceutical industry, and enhancing business processes is massive. Below, I discuss the most urgent steps the Generative AI industry should take to make the technology potent in the digital economy.

Increased Investment will improve accuracy and overall innovation

One of the biggest drawbacks in the Generative AI landscape is the technology's inaccuracy (commonly referred to as hallucinations) and malleability to bias. These issues are not just a matter of fine print disclaimer in the technology's terms and conditions but considerable drawbacks to the integrity of the platforms they run on - like ChatGPT, Bard, or Dall-E 2. Since this technology is perceived as a digital enabler in the 'knowledge economy', the promotion of accuracy and elimination of bias is at the core of their functional principles.

In the last six months of its' hype, Generative AI platforms across many companies have undergone rapid cycles of iterations. These iterations have been performed primarily to improve the accuracy of the LLMs that power the platforms.

OpenAI launched ChatGPT on November 30, 2022, with GPT-3.5 (GPT, Generative Pre-trained Transformer is a multimodal large language model that powers OpenAI's generative platforms). GPT-3.5 was upgraded to GPT-4 and launched on March 13 representing a 40% improvement in accuracy and cut-back on hallucinations. Google launched its LLMs called Med-PaLM on December 26, 2022, and later upgraded to LaMDA which powers its' Bard platform. Meta followed the train by launching its model, LLaMA, on February 24, 2023. Other notable players include Amazon, Microsoft, Cohere, Salesforce, Bloomberg, and Anthropic which have developed their own LLMs in just the last 5 months. All this development work has been enabled by the huge investments made to launch and iterate the LLMs and their interfaces.

About US$1.4 billion is estimated to have been injected into Generative AI projects in 2022. This is projected to reach US$11.3 billion for the whole of 2023. These iterations are currently largely focused on improving the language models and reducing hallucinations. However, a path to businesses adopting Generative AI will include customized innovations for corporate use cases. A good example of this customization is when Microsoft integrated GPT-4 into its' Office 365 suite and released this on March 16, 2023. Further investment and development work will create more potent enterprise use cases that will create a more robust commercial proposition for the technology providers. I will touch on these use cases in the sections below.

Microsoft announced an investment of US$10 billion in OpenAI in January 2023. More investment is required going forward to enable iterative improvements in the language models, extend use cases as well as support AI safety research and regulation. This funding is critical in enabling a path to corporate play in the Generative AI industry, which will in turn grow commercial outperformance for the platforms. The only downside I see to the push for funding is that the commercial models for Generative AI platforms are still largely untested. ChatGPT and MidJourney have been charging users monthly subscription fees of up to US$20 and US$60 respectively. It is still early days to say whether this model will grow and sustain commercial viability despite the massive user base so far amassed by ChatGPT.

A collaborative approach to AI safety and regulation will be critical for acceptance and growth

Generative AI has catalyzed the world's fear of 'the rise of robots'. This is not just blind apprehension, AI leaders themselves have expressed a rather unprecedented honesty about the impending danger of (Generative) AI and the risk to humanity.

In a government committee hearing over two weeks ago (May 16, 2023), OpenAI CEO Sam Altman was questioned about the potential dangers of Generative AI to which he responded,

My worst fears are that we (the AI field, technology, industry) can cause significant harm to the world...I think if this technology goes wrong, it can go quite wrong.

Of course, the implied dangers of the technology include,

• Algorithmic biases - if the model is trained on biased data the outputs will only reinforce bias, misinformation, and prejudice.
• Inaccurate or misleading information - this may have far-reaching consequences if applied to sensitive fields such as in health and pharmaceuticals.
• Disruption of intellectual property rights and plagiarism - since the LLMs use 'world-wide-web' data, they do not acknowledge its' provenance or the original creators.
• Cyber criminality - the development of realistic digital artifacts by the Generative AI technology may reinforce cyber attacks. This also includes using technology to acquire knowledge for harmful ends like terrorism.
• Disruption of business models and jobs - the technology is built on search and knowledge foundations. This will not only potentially disrupt industries whose models are built on information management and advertising but will also disrupt knowledge workers.
• The anthropomorphization of AI - the potential for the technology to take a life of its own and replicate beyond human-in-the-loop operations. We do not yet fully understand how this will impact us.

Earlier on in May 2023, Geoffrey Hinton - widely known as the 'Godfather of AI' quit his role at Google and stated in a tweet that,

I left so that I could talk about the dangers of AI without considering how this impacts Google.

The posture taken by AI leaders like Altman, Hinton, and others to raise concerns over the technology is clear enough. Now is the time for AI enterprises, governments, and other authorities to carefully craft and implement robust AI regulations. This regulation will not only manage the AI risks involved but will also serve to allay fears and encourage enterprise adoption of the technology. Without a solid regulatory blueprint and roadmap, enterprise leaders will be hesitant to adopt Generative AI. The banning of ChatGPT in Italy (and this consideration for TikTok in the USA over questions about its algorithm and user data protection), is a testament to how important it is to establish clear safety regulations.

Some of the most important elements of AI regulation will include algorithmic transparency, governance, and oversight. The emerging Generative AI platforms should collaborate with private organizations and authorities to create management frameworks for the algorithms involved. A multilateral oversight of the underlying algorithms will create a shared responsibility for AI safety, and entrench accountability for the AI players.

The establishment of an AI industry consortium, with proper authority, will be important in establishing overarching policies, rules, and regulations. While governing the industry, this consortium will be close enough to the technical development of the technology. This will create a symbiotic relationship that, unlike other industries, will not leave regulation lagging.

Another important step is for the Generative AI companies to fund research in AI safety. AI technology development is one thing; AI safety should be considered technically related but different. Significant funding for AI safety will thus enable prioritization.

Generative AI as a Service (G-AIaaS) will unlock the next wave of enterprise digitalization

The current scope of Generative AI platforms largely focuses on providing consumer interfaces for user experimentation. This is fine because it's still early days for the technology, and mass consumer onboarding is a great start in building equity. But a swift pivot to customized enterprise applications will unlock massive value for Generative AI companies. We have already seen some enterprise use cases, like the integration of GPT-4 into its Office 365 suite, GitHub Copilot for code generation, and Cohere Generate for commercial business operations. Such applications and other more specific technical use cases will unlock B2B opportunities.

The AI as a service (AIaaS) industry has already provided off-the-shelf AI tools that have enabled enterprises to implement and scale AI-enabled operations faster and at lower costs. These services include chatbots and virtual assistants, cognitive APIs, AI frameworks, and various AI-managed services. For many companies, AI applications are at the core of their digitalization and yet remain a black box. The development of more industry-specific AIaaS use cases in the Generative AI context will enable wider adoption of LLMs for enterprise use. Can we already start calling this G-AIaaS?

I believe some important G-AIaaS applications will develop new digital products, design new drugs, re-engineer business processes and improve supply chains. It is unlikely that many companies will have the funding to finance their own Generative AI projects. Therefore, the development of G-AIaaS propositions by Generative AI companies will be at the core of enabling digitalization from this technology.

In conclusion

Generative AI companies like OpenAI and Midjourney have already launched and shown their POC. The success in achieving over 100 million users for ChatGPT has unlocked massive value for OpenAI including a staggering funding round of US$10 billion from Microsoft this year. The next steps for the industry are critical in unlocking sustainable value for the business world.

More funding is required to continue the steep path of iterative improvements, AI safety needs to be systemically established both internally and externally, and more robust as-a-service use cases need to be developed. These elements need to be unlocked contemporaneously as the technology is already live.

The hyperbole about Generative AI some day curing cancer or resolving climate change will not create commercial sustainability. Like blockchain's fad, it is easy for any technology to fail at living up to its' full promise. The AI players must critically assess and execute on its' most urgent industry issues; funding, safety, and use cases in order to unlock value for the business world.

References & Notes

1. Digital natives (in this context) refers to companies established with a core digital operating model; with a digital technology platform and processes.
2. ChatGPT fastest growing consumer platform - https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/