9 ways to get started on AI today

Tech hype cycles can be exhausting. Remember the excitement about how everyone is going to have a 3D printer in their home? Then some cycles kept on coming back like crypto that had multiple lives from Bitcoin to ICO’s to NFT’s.

Is AI another one of these tech hype cycles? That’s a question I often hear from startup founders not already aboard the AI train. On the one hand, AI has had past hype cycles like when startups touted being “AI-driven” when in reality they were driven by mechanical turk.

Having experienced many of these cycles though, it is clear the AI hype of 2023 is real. The pace of innovation and investment energy around AI over the past year has been building significantly. In the most recent Y Combinator cohort, 94 startups (or 34%) listed themselves as AI companies, double previous cohorts.

But what does it mean to use AI? For startups with existing products, AI does not seem to be something that can be readily incorporated. But with the speed by which some AI-first startups are launching across every conceivable vertical and industry, many are founders are scared of getting left behind by these fast-moving AI startups powered by the likes of ChatGPT.

Using AI to stay competitive does not mean scrapping what you are building today. What has changed for startups though is that not using AI actively in some manner of the business will be a liability for future growth. Let’s take software development as an example where Paul Graham tweeted:

We used to say that 10X developers were rare like unicorns. What happens when AI tools enable nearly every developer to become a 10X developer? How does this translate into shipping new products and features? How does this change your time to market and ability to achieve product-market fit?

It can seem daunting however to figure out where to start with AI. How can you as a founder exploring the possibilities of AI actually start using AI to stay current and competitive?

We have broken out nine ways that startups are incorporating AI, categorized in three buckets from being interwoven to their product offering (Core), delivered as a service to customers (Application), or built into how the business operates (Applied).

Core AI

Data Layer – All AI is driven by machine learning algorithms built from data, particularly public data that drives Generative AI. The next step will be to inform LLM’s with private data sourcesc, which is sparking interest in synthetic data startups such as Rendered, Datagen, and Gretel that fill the gaps from public data sources and lack of access to proprietary data sources inside private companies.  

Infrastructure Layer – AI also needs plenty of compute to drive the initial training that allows AI to make useful decisions, as well as the ongoing inference when AI is tasked with making a real-time decision. As an example, AWS has specialized AI compute instances AWS Inferentia and AWS Trainium for inference and training workloads. This layer also includes everything that supports the heavy compute needs of AI, from energy management to security to data storage and many other additional services down to the chip level. And the compute needs to run at scale, which is where startups such as Anyscale come into play, making is easier to run distributed compute by orchestrating every aspect of the ML pipeline.

Foundation Model – This is the layer of the AI stack that has been getting all of the attention. While Open AI is the most well-known through Chat GPT, others are AI21 Labs, Anthropic, Stability AI, and Hugging Face as well as newer LLM’s such as Falcon LLM from the UAE. More models will continue to emerge for specialized needs, especially open source models that will be adopted by companies and governments that need AI but also have concerns about privacy, security, and IP.

ML Operations – To make AI work effectively requires a significant level of orchestration across data preparation and ingest, development, training, testing, deploying, inference, and more. MLOps manages this complex orchestration so companies can go from idea to shipping live AI applications. This has been an active space for startups prior to the Generative AI given how challenging deploying ML models can be, which is where startups like Abacus AI, Fiddler, Lightning AI, OctoML, Weights & Biases, WhyLabs, and others can help, as well as cloud compute services like Amazon SageMaker.

Application AI

AI Apps – This has been the most active area of startup formation in the past several months as more and more startups launch AI focused applications or pivot their startups to build on top of Foundation Models and plugging into tools like ChatGPT. If there was ever a time where the phase “there’s an app for that” was appropriate, it’s now as AI startups are launching across every conceivable industry and use case and business model.

Autonomous Agents – While most of the focus on application development has been on customer facing use cases, there is an emerging category of agent-based apps that work behind the scene to observe and make decisions, particularly in areas that are away from the human eye and involve massive amounts of real-time data such as agriculture, supply chains, logistics, and manufacturing.

Applied AI

ML-enhanced Data – Many startups have been dabbling in data science to start unlocking insights from their data to create better services or enhance the customer experience. Companies like Love Bonito and Foursquare are just a few examples of scaleups in ecommerce and geospatial service markets that have used AI extensively to bolster the value they bring to customers.

AI-driven Operations – There is no domain that is safe from being disrupted by AI. From marketing & sales to customer service to HR & finance, AI is enabling greater automation and efficiencies. Just in the realm of content marketing, there are apps that do the writing, create the images & videos, optimize the SEO, and automate the generation of social media snippets across every platform. This is an opportunity for startups to automate, accelerate, and scale aspects of their operations without the need to added resources or the cost of outsourcing this work.

AI-driven Development – As mentioned in the Paul Graham quote, AI applied to software development is a game changer for productivity that is bigger than even the advent of sites like Stack Overflow were, because now AI can let developers of any level of expertise create efficient working code faster than ever before. Tools like GitHub Copilot and Amazon CodeWhisperer are speeding the work of developers as smart code assistants while Amazon CodeGuru Security and DevOps Guru automate the management of security and DevOps orchestration in the code pipeline.

Many thanks to Jeremiah Owyang for his work on the AI Tech Stack that informed this post. And as you look through these nine ways to dive into AI, where do you see your startup using AI today?

Mark & Basil

Of the flood of AI related posts and content that I have been reading, one of the most informative was a post by Peter Yang, author of the Creator Economy newsletter, who shared this AI landscape in a recent LinkedIn post that showed significant fund raises by Generative AI startups across categories. Let us know if you find any other useful posts that are tracking this very active space!

We are super excited to officially launch our new YouTube channel also called Founders in the Cloud! This is a space to feature startup founders in short interviews to learn what they are building, how they went from idea to solution, and what they have learned so far in their startup journey. As we grow the channel, you will also hear commentary from us on startup trends, useful updates about AWS, and our journeys exploring various startup communities around the globe.

Our first series are interviews from the Startups Red Carpet during the Tel Aviv AWS Summit with 17 Israeli startups that we spoke with about their startup that are innovating across AI, cybersecurity, podcasting, and many other areas.

Tune in, let us know what you think, and please subscribe. If you have an interesting startup story to share, let us know!