Want to stay on top of artificial intelligence trends? Sign up to get regular updates and exclusive insights on AI that’s leading the industry.
Everyone seems to be talking about Generative AI these days. It’s a game-changer, promising to revolutionize everything, right down to our daily lives. But despite the buzz around it, a report from Menlo Ventures, which we got a sneak peek at, suggests that the excitement might be a bit overblown. Turns out Generative AI makes up less than 1% of what businesses spend on cloud services. That’s tiny compared to the 18% that goes to traditional AI out of the whopping $400 billion cloud market.
Derek Xiao from Menlo Ventures mentioned that there were big expectations for Generative AI to quickly take the world by storm. While AI is undoubtedly a huge leap forward, these things don’t happen overnight in the business world.
In 2023, spending on traditional AI continued to rise. Some experts are estimating that by 2030, Generative AI could be worth a jaw-dropping $76.8 billion, while others believe that it could generate at least $450 billion across various industries within the next seven years.
ChatGPT, since its launch in November 2022, has become a hot topic everywhere, from the boardroom to casual water-cooler chatter. Before 2023, half of the companies reported using some form of AI, and adoption has increased since then. The investment in AI across these businesses has gone up about 8%, with product engineering departments being the top spenders.
However, companies are still cautious about diving into Generative AI. Naomi Ionita of Menlo Ventures noted that the past year felt more like a trial phase rather than full adoption.
Looking to the future, 2024 looks set to be the year of serious Generative AI implementation, according to Xiao. But company leaders can take heart; it’s perfectly fine to take things slow, Menlo’s Tim Tully suggests. Given how rapidly Generative AI is evolving, many are hesitant, and often the funds just aren’t there yet. After all, investing in new tech is a significant financial decision.
The journey of adopting new tech, like the cloud in its early days, is expected to be gradual. The most significant hurdles? Proving a return on investment and the “last mile problem,” where the final step of implementation proves the trickiest. Add to that the headaches around data privacy, the ongoing search for AI talent, getting the new tools to fit with the old systems, and the need for some tech that explains itself better and can be customized.
The report also indicates that current enterprise solutions haven’t quite lived up to their promises of dramatically transforming business operations. Until businesses see real value, many will remain skeptical. And it doesn’t help that it’s become tougher to get spending past the chief financial officer.
However, there are early adopters who are already benefiting from Generative AI, notably in data management and eliminating time-consuming tasks. This is breaking new ground in user experience, a fact echoed by Ionita. Tully even mentioned that new and impressive tools can be whipped up in 20 minutes or less, making jobs easier and fostering success.
Menlo Ventures is hopeful for both industry-specific and more general applications of Generative AI. For instance, marketers are getting a lot out of Synthesia for creating video content, while the legal sector has Harvey for checking contracts. Then there are tools designed to simplify routine tasks. Looking ahead, expect AI tools that think and act more independently, taking on things like email, scheduling, and note-taking, freeing up time for employees currently juggling multiple applications.
Moving forward, the novelty of AI will wear off, and it will become an integral part of the workday. As for the technical side of things, companies pumped $1.1 billion into the modern AI stack, with closed-source models like those offered by OpenAI and Anthropic dominating the market. Yet, the majority of businesses are still picking off-the-shelf models over custom pre-trained ones.
Companies are looking for models that offer more control and cost savings, with most of their AI budget going towards running the AI, not training it. Popular tweaks include adjusting prompts for better results, while human review remains a favorite for assessing performance.
An emerging standard is retrieval-augmented generation, which boosts large language models with external data, ensuring responses stay fresh and relevant. Enterprises are also dabbling with fine-tuning, adapters, and reinforcement learning to make AI even smarter.
The landscape for the first part of the year was a bit of a free-for-all, but the industry is starting to agree on key components and practices. Despite this, the AI stack isn’t yet a cookie-cutter solution, which spells opportunity for startups. There’s a need for services to manage and deploy models remotely, as well as tools for creating data pipelines and ensuring content governance.