If you can, think way back to 2008 when Apple first launched its App Store. At its inception, there were a fraction of the downloads we have today as consumers started to adopt a new way to work, play and travel and as developers learned new ways to create impactful products for end users.
We’re in a new yet similar era of innovation at present.
With the rise of generative AI tools entering the market, we’re just now scratching the surface regarding what’s possible. So far, generative AI has played in the platform layer of a tech stack: with code occurring in your integrated development environment or images being curated in MidJourney and iterated on in Canva, Figma or Photoshop. There are only a handful of standalone web apps, like Jasper and Copy.ai, in the marketplace today.
With so much attention at the platform layer, the application layer is ripe for disruption. Just as app developers thrived during the heyday of mobile innovation, we expect the new large language models (LLMs) to motivate a new wave of generative AI applications. Later, applications will be replaced with AI-native workflows.
Currently, and likely for the foreseeable future, generative AI frameworks will continue to be dominated by large companies with deep pockets. The time, cost, and expertise associated with running and maintaining the large clusters of graphic processing units (GPUs) or tensor processing units (TPUs) required to train billions of parameters as needed for LLMs, multimodal models, et al., make it challenging for small companies or solopreneurs to play in the space.
Luckily, open-source models provide implementations that anyone can freely use and modify, and the wide availability of low-cost, hosted models makes access to these technologies more possible every day.
Therefore, the opportunities for innovation and disruption lie once training of the foundational model is complete. Businesses can then fine-tune the foundational models for specific, niche use cases, addressing particular problems in areas like code, design, and gaming, and run these unique models within their applications.
Those who embrace fine-tuning foundation models, or solving a specific issue, as opposed to being everything for everyone, can expect to achieve the greatest differentiation and competitive advantages.
We also think this approach will shape an entirely different type of business ecosystem. While there will always be competition, there is now more space for collaboration, networking, and community.
Democratizing data across a tech stack has been a significant messaging point for organizations for years. Bringing together data across disparate data systems remains one of the biggest challenges for applications, as well. Still, the more comprehensive and actionable a data set, the more productive, faster, and smarter an organization becomes. Succinct tech stacks often translate into better customer experience for the end user, a point of high importance for modern brands.
With the continued adoption of generative AI, we believe customer demands for customization and personalization will continue to increase, and we’ll finally have the technology to keep pace.
To solve a customer’s needs, we anticipate democratization of data will now extend beyond any individual organization and may soon mean pulling information across a broader ecosystem.
End users will continue to value simplicity and convenience. And, leaders tapped into these preferences will be able to push personalization to heights we’ve never seen before; tailoring all aspects of digital interaction to how the customer wants it to flow- from the first touch to the last - even if it means linking to outside partners.
For example, let’s say someone is considering a home renovation project. What if they could type a query into a single text prompt asking:
“I have a $5000 budget. Here are seven photos of my current kitchen, and seven inspiration images I want to achieve. Now give me a complete renovation plan, including design, and put in bids for me."
Even if the results require human-in-the-loop iterations, the consumer can still accomplish an extremely customized project quickly and with fewer resources.
While this is not currently an option presented to the marketplace, we believe by combining proprietary data and fine-tuned models with public datasets, and across other AI tools - including multimodal AI options - we can improve AI’s ability to understand context, predict what is being asked.
With a broader pool to execute a command, we think a single text prompt like “What do you want to do today” will soon be a full-service query.
As new techniques shrink the costs required to train better and larger models, and as developer access expands from closed beta to open beta or open source, innovation will skyrocket. At the same time, businesses will continue to turn to AI to meet consumer demands.
As these two individual initiatives inevitably intersect, everything we thought we knew about business and the ecosystem will change and become more democratizing.
Curious to see how Evolv AI plays a role in the larger ecosystem? Connect with us to learn more.