Hype cycles are intense. The mainstreaming of transformative technology inspires innovators and imitators alike, bringing forth grafters and grifters in equal measure. For those on the outside looking in, it can be hard to separate signal from noise; the startups that are changing the world from those that are just changing lanes.

This is particularly true at the intersection of AI and web3, two genuinely novel technologies, each with the power to dramatically alter the way we work, communicate, transact, and recreate. It’s easy to stuff a whitepaper with buzzwords, ride the wave, and take the masses along for the ride. It’s harder to build a product that actually works.

30 years ago, the internet existed on a desktop in your home study. Today, it’s everywhere, connecting your heating, home, car, fridge, TV, wristwatch, and pet’s collar. Something similar is now happening with AI, which is permeating industries and upending everything from video to fashion.

The AI arms race has seen demand for GPU compute elevate Nvidia into the third largest company in the S&P 500. The OpenAI-partnered Microsoft, meanwhile, has reclaimed its top spot from Apple. Investment is pouring into artificial intelligence right now, whose CAGR stands at 37% per year.

Companies of all kinds are currently creating products that utilize AI or provide the infrastructure for AI startups. There’s now a thriving industry centered around supplying the GPU compute required to operate large-scale AI models. Demand for GPUs is increasing by 10x every 18 months, while GPU capacity is only doubling in the same amount of time. The AI express is gathering speed. But not all of the companies aboard are contributing to its success; some are merely along for the ride.

It’s common to see startups claiming to create the AI platform for everything. Training, inference, compute: you name it, they purport to do it. While such projects may have good intentions, they have a long roadmap ahead of them which will require high amounts of resources, expertise, and iteration to deliver on their promises. Building a product that works takes time, tooling, and real users. It calls for knowing your target market and providing them with a solution that’s usable in time to make a business impact for them – not in some distant and undefined future.

Livepeer has focused on developing practical solutions that solve problems for video-focused developers today looking to build on AI capabilities, starting with the availability of GPU compute. We see particular value in supplying the tooling to support inference: using trained AI models to produce outputs or make predictions based on the set of inputs. This is a task ideally suited to a decentralized network such as Livepeer, with node operators able to load specific models onto their GPUs and compete on the cost to perform inference jobs.

The Livepeer network is striving to handle a broad range of AI video compute tasks, such as generating video from text – as Sora has so powerfully demonstrated recently. We’re now constructing the tooling that will power a range of job types including AI-based generative video. This will result in a full stack of video solutions that work for such tasks as ingestion, storage, delivery, and transcoding.

AI is going to change the world and we’re confident that Livepeer will play a role in that seismic shift – with no vaporware, no hype, and no buzzwords for buzzwords’ sake. Instead, we intend to ship practical solutions that solve real-world problems. We’re here to make a difference and we’ll still be here long after the hype has died down and AI is everywhere.

Hype cycles are intense. The mainstreaming of transformative technology inspires innovators and imitators alike, bringing forth grafters and grifters in equal measure. For those on the outside looking in, it can be hard to separate signal from noise; the startups that are changing the world from those that are just changing lanes.

This is particularly true at the intersection of AI and web3, two genuinely novel technologies, each with the power to dramatically alter the way we work, communicate, transact, and recreate. It’s easy to stuff a whitepaper with buzzwords, ride the wave, and take the masses along for the ride. It’s harder to build a product that actually works.

30 years ago, the internet existed on a desktop in your home study. Today, it’s everywhere, connecting your heating, home, car, fridge, TV, wristwatch, and pet’s collar. Something similar is now happening with AI, which is permeating industries and upending everything from video to fashion.

The AI arms race has seen demand for GPU compute elevate Nvidia into the third largest company in the S&P 500. The OpenAI-partnered Microsoft, meanwhile, has reclaimed its top spot from Apple. Investment is pouring into artificial intelligence right now, whose CAGR stands at 37% per year.

Companies of all kinds are currently creating products that utilize AI or provide the infrastructure for AI startups. There’s now a thriving industry centered around supplying the GPU compute required to operate large-scale AI models. Demand for GPUs is increasing by 10x every 18 months, while GPU capacity is only doubling in the same amount of time. The AI express is gathering speed. But not all of the companies aboard are contributing to its success; some are merely along for the ride.

It’s common to see startups claiming to create the AI platform for everything. Training, inference, compute: you name it, they purport to do it. While such projects may have good intentions, they have a long roadmap ahead of them which will require high amounts of resources, expertise, and iteration to deliver on their promises. Building a product that works takes time, tooling, and real users. It calls for knowing your target market and providing them with a solution that’s usable in time to make a business impact for them – not in some distant and undefined future.

Livepeer has focused on developing practical solutions that solve problems for video-focused developers today looking to build on AI capabilities, starting with the availability of GPU compute. We see particular value in supplying the tooling to support inference: using trained AI models to produce outputs or make predictions based on the set of inputs. This is a task ideally suited to a decentralized network such as Livepeer, with node operators able to load specific models onto their GPUs and compete on the cost to perform inference jobs.

The Livepeer network is striving to handle a broad range of AI video compute tasks, such as generating video from text – as Sora has so powerfully demonstrated recently. We’re now constructing the tooling that will power a range of job types including AI-based generative video. This will result in a full stack of video solutions that work for such tasks as ingestion, storage, delivery, and transcoding.

AI is going to change the world and we’re confident that Livepeer will play a role in that seismic shift – with no vaporware, no hype, and no buzzwords for buzzwords’ sake. Instead, we intend to ship practical solutions that solve real-world problems. We’re here to make a difference and we’ll still be here long after the hype has died down and AI is everywhere.