Industry executives and experts share their predictions for 2024. Read them in this 16th annual VMblog.com series exclusive.
Generative AI Transforms Coding (and Gets Practical)
By Shomron Jacob, Head of Applied Machine Learning and Platform at Iterate.ai
The breakneck pace of generative AI’s evolution-and its subsequent impact on enterprise technology practices-will continue to accelerate in 2024. Coding and application development are particularly ripe for AI transformation, and code-generating LLMs that have just now crossed crucial thresholds in their sophistication will only become more powerful going forward. At the same time, expect more enterprises to recognize the value of nurturing their own private LLMs and more mature generative AI strategies in the coming year-and those that get this right will break away from competitors.
Let’s take a closer look at the changes these trends will bring in 2024:
1. Generative AI-based code-generation LLMs will offer unmatched business value.
The hype over generative AI will settle down to reveal the technology’s most practical and effective enterprise use cases in 2024-with the ability to deliver fully functional and usable code offering perhaps the biggest night-and-day advantages over legacy processes. Organizations that pursue code-generation LLMs (we recently launched our own, Interplay-AppCoder) will equip their developer teams with far greater productivity, resulting in faster product and customer experience improvements and more compelling innovations.
Developers once slowed by tedious block-and-tackle coding will see many of those burdens automated away, freeing them to focus on tasks with greater strategic and competitive value. Additionally, rapidly advancing code-generation LLMs-which already provide unprecedented efficiency and agility for development teams-will be far more powerful by the end of 2024 than they are at the beginning, offering increasing competitive benefits as the year progresses.
2. Low-code pairs with code-generating LLMs to transform developer and DevOps productivity.
The acceleration that low-code development strategies currently achieve for development and DevOps teams will shift into a whole new gear in 2024, as low-code code-generation tools arrive to push the envelope on efficiency. In the right hands, these tools will knock down some of the final barriers when it comes to empowering developers and DevOps to execute their visions and seamlessly deliver applications with innovative capabilities. The competitive gap between products and customer experiences offered by enterprises that utilize low-code code generation and those that don’t will be increasingly stark over the coming year.
3. Enterprises that invest in private LLMs will control their own futures.
Organizations have a decisive choice to make in either relying on available public LLM APIs like OpenAI, or building and training their own private LLMs. While training a private LLM requires investments in computing power, doing so allows an enterprise to retain its own private data and build its own IP that it fully controls.
Many organizations will choose the easier path of utilizing public LLMs, and their applications and features will feel more commonplace as a result-perhaps to the point of seeming mundane as more and more organizations harness the same tooling. In contrast, organizations with private LLMs will have capabilities as powerful and unique as their proprietary data enables, driving distinctive experiences and clear competitive differentiation. As practices mature, opting for a private LLM versus a public one may be the single biggest determining factor in whether enterprises are able to harness generative AI’s full potential.
4. LLM utilization will become more strategic.
Again, LLM hype will give way to practical concerns in 2024, allowing organizations with more prudent strategies to stand apart. For example, traditional AI is more appropriate and effective than LLMs in many use cases. That’s especially true when it comes to query response times and efficient usage of computing power.
However, it will take clear strategies and restraint to apply the right AI tools to the right tasks. Addressing LLM-based risks-with AI hallucination as a primary concern-will also be a strategic priority. Enlisting traditional AI to handle sensitive tasks such as financial transactions (trading unique customer experiences for predictability) will be the better choice in certain high-risk circumstances.
5. Contextually-aware interactive AI will emerge, moving capabilities a step beyond generative AI.
AI can be broken down into four categories-classification, generative, interactive, and general-which provide a useful framework for understanding AI’s evolution. Classification AI is the most basic form of AI, where a system classifies data into predefined categories, as seen in common applications such as image recognition, spam filtering, and recommendation systems. Generative AI goes beyond classification by creating new data that resembles input data, enabling today’s solutions like GPT, DALL-E, and others.
We’re now rapidly progressing in the generative AI category, with breakthroughs in emerging interactive AI just around the corner in 2024. Interactive AI, represented by GPT-3 and GPT-4, engages in two-way interactions with users and provides dynamic responses that are increasingly contextually aware (and responsive to) users’ needs. That crucial contextual awareness makes the difference between simply reacting to user input data and queries, and shaping results with a deeper understanding of what the user requires as an individual at that moment.
The fourth phase, general AI, refers to AI that can understand, learn, and apply intelligence as broadly and flexibly as a human. While those capabilities are certainly farther on the horizon, the contextual awareness now emerging with interactive AI is a key milestone in reaching that future.
##
ABOUT THE AUTHOR
Shomron Jacob is the Head of Applied Machine Learning and Platform at Iterate.ai, whose AI innovation ecosystem enables enterprises to build production-ready applications. Shomron began his career as a software engineer but soon found himself learning ML/AI, and switched his professional direction to follow it. He lives in Silicon Valley.
