AI tech startups may face a financial nuclear winter

Couldn’t attend Transform 2022? View all summit sessions in our on-demand library now! Watch here.


Recently, I wrote a piece for VentureBeat distinguishing between companies that is They are based on AI at their core and those that only use AI as a function or a small part of their overall offering. To describe the former set of companies, I coined the term “AI-Native.”

As a technologist and investor, the recent market downturn got me thinking about the technologies poised to survive the AI ​​winter — caused by a combination of reduced investment, temporarily depressed stock markets, a potential recession exacerbated by inflation, and even from customer hesitancy about dipping their toes into promising new technologies for fear of missing out (FOMO).

You can see where I’m going with this. My view is that AI-Native companies are in a strong position to emerge healthy and even grow from a downturn. After all, many great companies have been born during periods of downtime – Instagram, Netflix, Uber, Slack, and Square are a few that come to mind.

But while some unheralded native AI company could become the Google of the 2030s, it wouldn’t be accurate—or wise—to proclaim that all AI companies are destined for success.

Event

MetaBeat 2022

MetaBeat will bring together thought leaders to provide guidance on how metaverse technology will transform the way all industries communicate and do business on October 4 in San Francisco, California.

Register here

In fact, native AI companies have to be especially careful and strategic in how they operate. Why; Because running an AI company is expensive — the talent, infrastructure, and development process are all expensive, so efficiency is key to their survival.

Want to tighten your belt? There’s an app for that

Efficiency isn’t always easy, but luckily there’s an AI ecosystem that’s been around long enough to provide good, useful solutions for your particular technology stack.

Let’s start with model training. It is expensive because the models are getting bigger. Recently, Microsoft and Nvidia trained the Megatron-Turing Natural Language Generation (MT-NLG) model on 560 Nvidia DGX A100 servers, each containing 8 Nvidia A100 80GB GPUs — costing millions of dollars.

Fortunately, the cost is coming down due to advances in hardware and software. Both algorithmic and systems approaches such as MosaicML and Microsoft’s DeepSpeed ​​create efficiencies in model training.

Next is data labeling and development, which [spoiler alert] it is also expensive. According to Hasty.ai – a company that aims to tackle this problem – “data labeling takes anywhere from 35 to 80% of project budgets.”

Now let’s talk about creating models. It’s hard work. It requires specialized talent, a ton of research, and endless trial and error. A big challenge with building models is that the data is environment specific. There has been a post for this for a while. Microsoft has Azure AutoML, AWS has Sagemaker. Google Cloud has AutoML. There are also libraries and collaboration platforms like Hugging Face that make creating models much easier than in years past.

Not just releasing models into the wild

Now that you’ve created your model, you need to develop it. Today, this process is painstakingly slow, with two-thirds of models taking over a month to develop into production.

Automating the development process and optimizing for the wide range of hardware and cloud service targets supports faster innovation, enabling companies to remain ultra-competitive and adaptable. End-to-end platforms like Amazon Sagemaker or Azure Machine Learning also offer deployment options. The big challenge here is that cloud services, endpoints and hardware are constantly moving targets. This means that new iterations are released every year and it is difficult to optimize a model for an ever-changing ecosystem.

So your model is now in the wild. And now what? Sit back and put your feet up? Think about it again. Models are broken. Continuous monitoring and observability is key. WhyLabs, Arize AI, and Fiddler AI are among the few industry players tackling this issue head-on.

Beyond technology, talent costs can also be a barrier to growth. Machine learning (ML) talent is scarce and in high demand. Companies should rely on automation to reduce reliance on manual ML engineering and invest in technologies that fit existing application developer workflows so that more DevOps professionals can join the ML game.

The native AI company: Solving all these elements

I would like to add a suggestion about agility/adaptability. If we’re talking about surviving a nuclear winter, you’ll be the most hyper-competitive and adaptable — and what we’re not talking about here is the real lack of flexibility when it comes to ML development. The automation we bring is not just the adaptability part, but the ability to innovate faster – which is currently limited by incredibly slow development times

Fear not: artificial intelligence is coming of age

Once investors have spent their time and paid some fees (usually) in the venture capital world, they have a different perspective. They have experienced cycles playing with technologies they have never seen before. As hype takes hold, investment dollars pour in, companies are formed, and new product development heats up. It is often the quiet tortoise that eventually wins over the investment rabbits as he humbly gathers users.

Inevitably there are bubbles and busts, and after each crash (where some companies fail) the optimistic predictions for the new technology are usually exceeded. The adoption and popularity is so widespread that it just becomes the new normal.

I have great confidence as an investor that regardless of which individual companies dominate the new AI landscape, AI will achieve much more than a foundation and unleash a wave of powerful intelligent applications.

Luis Ceze is a venture partner at Madrona Ventures and CEO of OctoML

DataDecisionMakers

Welcome to the VentureBeat community!

DataDecisionMakers is where experts, including data technicians, can share data-related insights and innovations.

If you want to read about cutting-edge ideas and up-to-date information, best practices and the future of data and data technology, join us at DataDecisionMakers.

You might even consider contributing an article of your own!

Read more from DataDecisionMakers

Leave a Comment