Top Five AI Predictions for 2024

It’s predictions time of year again. 2023 was unique in so many ways. We saw the worst technology job market since 2001, with hundreds of thousands in our industry still job hunting. Meanwhile, we’ve seen an unprecedented level of hype around Generative AI, including continued strong VC investments in early stage firms in that space, focused mainly Seed or Series A.?

What, then, will 2024 bring in technology? Below are my bets.

1 - Generative AI will move past peak hype and focus on use-cases and improving user experiences. Early adopters will have done enough prototyping and tire kicking to figure out how Generative AI can create real value through changing the way applications work to create interactive user experiences.?

Top GenAI Use-cases for 2024:

• Guided Coding & Code Reviews - there is a HUGE opportunity for GenAI to become an “Intelligent Stack Overflow” coders assistant to help find/create code snippets, search documentation, and accelerate coding. Generative AI can also catch many code issues so will reduce the amount of code review required. Given coders are high cost we can expect a strong focus on this among tool vendors and in the Open Source ecosystem.
• Conversational AI - Chat-bots and other interactive applications. OpenAI with ChatGPT has shown the world the power of GenAI and it’s what people think of first as a use-case. This can be a powerful tool when applied to offload common questions asked by customers making support/service calls and thus drive savings in call center staff costs.
• Interactive search - Remember “Ask Jeeves?” It was before its time but people do like the idea of a more interactive search. Many already ask questions to search engines. This use-case is also easy to integrate into mobile devices and into applications.
• Guided writing - we will see continued increase in GenAI in this area which was already a top use-case. As this gets integrated into content tools such as Google Docs, MS Word, tools like Wiki, etc., this will drive continued adoption. There will also be a backlash as it will become harder to know what bits were written by people or AI, and the lines will be heavily blurred. After all, we’ve been using specialized AI to check and fix grammar and spelling for decades, and most already use auto-complete generative language features today.
• Multi-Modal - this will be among the most hyped areas for Generative AI in 2024, as Big AI (from Google, Facebook, Microsoft/OpenAI, IBM, Amazon) as well as Open Source will be increasing focus on this area to drive more differentiation and drive a larger Target Addressable Market. While it will see huge hype, there is also huge controversy around Intellectual Property protections and abuses.?
• Creating graphics - text-to-image GenAI (and multi-modal) will see further growth, yet with continued controversy around how this relates to issues around Intellectual Property as well as related concerns over plagiarism

2 - Fine Tuning on specific use-cases and private data sets will overtake giant “Generalist” hosted Foundational Models.

Addressing “hallucinations”?and tailoring to vertical use cases and private data sets will bring Generative AI into new areas using fine tuned LLMs, optimized for specific use cases, or for vertical domains, such as Financial Services, Healthcare, Legal, etc. This will enable better experiences and higher quality outputs.

3 - RAG or Retrieval Augmented Generation architecture will dominate, leading to new innovations in the architecture. RAG will become the de facto pattern to address hallucinations, and for fine tuning to specific use-cases and training data. This will also enable more contextually accurate responses along with pointers back to source content.

4 - Edge & On-Device Inference will further drive demand for more sophisticated MLOps and model governance and enhanced SoCs that are optimized for these newer ML techniques. Privacy, and performance (low latency use-cases) will drive many inference use-cases towards deploy on-device or at Edge.

5 - Generative AI will be widely weaponized by bad actors.

We will see a number of high profile examples of where Generative AI will be leveraged by bad actors amongst cybercriminals and government sponsored groups and driven by bots. Top examples:

• Misinformation & Propaganda, Fake News, etc. - this will be everywhere, particularly by groups seeking to disrupt high profile elections and political events. We will see ever more realistic deep-fakes and some of these will end up inadvertently picked up by actual news media.
• Phishing & Scams - Similar techniques will be used to create ever more realistic phishing and scam content fed to us via Social, Text, Email and Web Sites.

Conclusion

2024 will be the year where Generative AI rubber meets the road. It won't be enough to ride the hype wave. Successful AI initiatives will focus on high value use-cases that can deliver quick wins. Firms that don't deliver value, and do so quickly, will hit a wall when the hype begins to fade. And those that haven't built solutions based on sound architectures that take care of security, privacy, governance, ML Ops and other challenges will struggle to achieve scale and repeatability with AI.