In the startup and VC community, we’ve been astonished by the rapid pace of innovation and advancement in AI technology in recent weeks.

GPT-4 was just released two weeks ago, and we’re already seeing companies scrambling to keep up with the new and improved language model.

The new plugins are especially impressive with features like language translation and sentiment analysis. Plugins are a game changer that will revolutionize the way companies interact with language models in the future.

Our very own Scott Moss has spent a good deal of time tinkering with Chat GPT-4 and exploring its capabilities. According to him, while the model’s ability to maintain context feels human-like, it has limitations when it comes to more creative tasks. Scott believes that Chat GPT-4 is at its best when used in a productivity context.

We reached out to several Initialized AI founders in our portfolio to get their first impressions of Chat GPT-4. Here’s what they had to say:

Pratap Ranade, CEO of Arena AI

Pratap Ranade, CEO of Arena AI, is impressed with innovations such as the human feedback component in GPT-4.

He’s particularly excited about the multi-model capability of GPT-4, which allows it to learn general representations of things that are portable between text and image.

“The idea that we are able to build machines that can learn general representations of things that are portable between text and image really feels like a profound step forward,” he said. 

Pratap also notes that while there is currently a boom in generative AI, it is part of a bigger and longer arc that has a ton of immediate applications, ranging from customer support to drug discovery. 

He sees generative AI as paving the way for entirely new classes of fully autonomous AI that can take input from multiple sensors (images, video, audio, and text) and create a meaningful representation of them.

“With generative AI, we can query this representation and see remarkable outputs,” he said. “But the next step would be chaining that together with other AI systems that can act, to create powerful closed loop autonomous systems.”

Jordan Fisher, CEO of Standard AI

Jordan Fisher, CEO of Standard AI, has been impressed with the accessibility that GPT-4 has brought to the world of AI. He notes that with just a team of three and a project manager, what used to take a large team years to build can now be accomplished in just weeks.

Jordan believes that this accessibility has enabled not only startups but also larger companies to become more agile in their approach to innovation. 

“In this new environment we’re seeing the larger incumbents leading the charge because the technology is so accessible,” he said. 

Jordan is particularly impressed with the speed at which new tools like GitHub Copilot X are being developed and brought to market.

The latest version of this generative AI coding tool, which features GPT-4 integration, has the potential to speed up coding by 55%. According to Jordan, this can change the dynamic and structure of many engineering teams.

With the integration of Chat GPT-4 into Copilot X, the need for human interfacing is reduced, enabling junior engineers to level up and grow while simultaneously allowing senior engineers to work more effectively. This not only saves time but also reduces the likelihood of errors that need to be corrected.

“The speed of adoption is like nothing I’ve seen before and it’s because the technology is so flexible and powerful,” he said.

Ivan Lee, CEO of Datasaur

Ivan Lee, CEO of Datasaur, said a story that’s not discussed right now is what AI means for economic equality around the world. Natural Language Processing and LLMs have made the most advances in English, Mandarin, and Spanish because the researchers primarily speak these three languages, according to Ivan. 

“The vast majority of the remaining 7000 languages in the world have not made even 10% of the progress of these languages,” he said.

While it’s clear OpenAI has now integrated “expertise” modules to help with some of the most common mistakes made by previous generations of GPT with the new plugins, Ivan is still worried about AI hallucinations.

“OpenAI has been forthcoming about this flaw — as a massive oversimplification, LLMs are just very advanced pattern recognition algorithms,” he said. “However, many of my friends have been turning to ChatGPT first over a traditional search engine like Google. There is a real risk we act on incorrect information.”

Still, our managing partner Brett Gibson recommends early stage startups to think about an AI strategy, even if it’s not necessarily core to the business at the moment. 

“It’s not going to be useful in everyone’s product, but startups should be paying attention because most tech companies have software components and it’s going to change the way software is written”.