Follow us on social media

Challenges and Contrasts for AI-powered Startups Between Open Source Models and ChatGPT

Challenges and Contrasts for AI-powered Startups Between Open Source Models and ChatGPT

In today’s tech-driven world, Venture Capital firms investing in startups can’t simply stick to funding as their primary -and only- value. Not even ‘smart money’, the term that applies to funding plus access to a pro network, is enough. In order to run the extra mile there has to be a full collaboration between investors, founders and boards. We call it ​​commitment: one of GoHub Ventures’ values expressed in our previous article ‘Anatomy of a Rebranding’.

So, in order to put our money where our mouth is, GoHub Ventures hosted its inaugural edition of Tech Unpacked, a series of events designed to explore cutting-edge technologies, foster networking and build key alliances.

 

Tech Unpacked #1: Summary

On a show conducted by Guillermo Flor (GoHub Ventures) under the headline of ‘Open Source AI vs ChatGPT: How to Build Privacy Aware AI Apps’, attendees heard expert insights from Iván Martínez (PrivateGPT) and Rory McElearney (Valencia Codes) about the considerations in data privacy when deciding whether to build or buy an AI-based solution. After listening to the experts, participants got down to work on a hands-on hackathon that elevated the experience into an interactive evening that ended with pizza for everyone.

To kickoff Tech Unpacked #1, several startups from GoHub Ventures’ portfolio introduced real cases of AI-powered solutions they’re currently serving to clients:

  • David Purón (Barbara) explained how Spanish multinational Acciona (a conglomerate dedicated to the development and management of infrastructure and renewable energy) has implemented AI to predict in real time the amount of chemicals their plants need for operations. Thanks to this, the company is more efficient when it comes to operational costs and also improves safety levels at their facilities.
  • Felix Laumann (NeuralSpace) detailed how oil giant Aramco uses their DocAI solution to analyze complex documents and extract qualitative content, a process previously done 100% manually.
  • Fernando Pérez (Aunoa) told how their AI Chatbots improve customer experience by automating interactions and conversations on WhatsApp, Facebook or a website.

 

Build or Buy an AI-based solution

Expert’s panel with Iván Martínez (PrivateGPT) and Rory McElearney (Valencia Codes) discussed the convenience of building or buying an AI-based solution. As in many cases, the answer is… it depends!

According to the invitees, one must think first about the problem they’re trying to solve. If the problem is core to business, building your own AI-based solution is a road you may want to explore…as long as you have the knowledge and legal aspects covered. However, if the problem isn’t core to your activity, you may want to take advantage of existing AI-based solutions rather than building your own.

When building your own AI-based solutions, infrastructure and privacy are another two aspects to take in consideration. First, you’ll need to host your model locally or cloud and this might get expensive so… you may want to think twice and follow the golden rule of ‘just build when it’s core’. Second, you become responsible for what data and how safely you store it… and legislation is different depending on where you’re based.

Existing AI-based solutions are perfect for try and error, since you can create magic in a few hours…as hackathon attendees proved. Teams of tech and non-tech profiles were formed on the fly and they just needed a couple hours to transform these ideas into MVPs:

  • Summarization solution for notes taken during working meetings.
  • Git commits message improver and tests code generator.
  • Psychologist assistant trained with psychology books.
  • Sales sentiment analysis using sales calls as input.
  • Marketing content generator using past posts.
  • User interface for PrivateGPT with add-ons like sessions, history, login and more.

At the current stage of AI’s development and adoption, the experts highlighted a few scenarios when AI is a perfect partner for companies:

  • Document Analysis & Summarization.
  • Customer Support.
  • Content Creation.
  • Business Intelligence (take better decisions with more data gathered with AI support).

 

Future of AI: super customization, fakes and jobs threat

Tech Unpacked is also about expressing views and forecasting what’s next, so attendees also had the chance to hear from experts about the future of AI.

Three industries where AI is going to have a massive impact, according to Iván Martínez (PrivateGPT) and Rory McElearney (Valencia Codes) are LegalTech, Fintech and HealthTech.

To understand what degree of disruption and how this can impact our lives, the experts explained a case within the Health sector. According to their opinion, we are walking into an era of super customization thanks to AI.

If you think of two patients today who have been diagnosed with the same illness, they’re probably getting a very similar treatment. However, they’re two very completely different persons. With AI’s penetration on the Health industry, the treatments that are similar today will tend to a higher level of customization.

Based on economic history, experts also agreed that manual low value work will go away while AI’s adoption increases, as it has happened through decades with any revolution whether it’s mechanical or technological.

Last but not least, experts also expressed their concerns about AI created content. According to their opinion, we’re not yet ready to distinguish real content from an AI created one. They consider this a threat that may affect from a simple domestic cybercrime to a massive fraud.

However, as the winners of Tech Unpacked hackathon showed, AI could be also super helpful for content creation. The group created a solution that produces content for marketing purposes (blog posts, shareables, etc.) by running AI models through previously published texts. A big help for marketers that today spend tons of time selecting and creating bullet points and catchy phrases for their audiences.

Who knows if soon they’ll become a new GoHub Ventures portfolio member!

Feb 01 · 2024 Cynthia Martínez Operations Manager