Testing AI applications for labs

Over the past few months, Coboi lab has joined the exploration and experimentation of the potential of new artificial intelligence (AI) applications. More specifically, we focused on testing those applications that could be beneficial in the day-to-day operations of a public innovation lab.

The starting point for our testing were two of the most popular “chats” currently available: ChatGPT (by OpenAI) and Bing (by Microsoft). We asked both the following question: “Which free artificial intelligence applications could be useful for a public innovation lab ? ”

From their respective answers, we compiled an initial list of recommendations:

• AutoDraw : A free AI app by a Google parent company that allows for easy and quick drawing.
• AIVA : Although the free version has limitations, it is fully available for use.
• DALL-E : This app is another one of the most well-known along with ChatGPT.
• TensorFlow : An open source machine learning platform used to build neural network models. It is one of the most popular and widely used AI tools in the world.
• Keras : A high-level machine learning library that runs on top of TensorFlow and allows you to quickly build neural network models.
• PyTorch : An open-source machine learning platform used to create neural network models. It is very popular among researchers for its usability and flexibility.
• Hugging Face : An open-source learning library used to build natural language processing (PLN) models and to create chatbots.
• OpenCV : An open-source library for image and video processing, widely use in computer vision applications.
• Caffe : An open-source deep learning framework used to build for building neural network models. It is often used in machine vision applications.
• Python : It is very popular due to its ease of use and its wide range of automatic learning algorithms.

However, ChatGPT itself (after sharing its first list of “top apps”) invited us to specify which needs (of labs) we wanted to adress with artificial intelligence applications, since the current offerings are already substancial:

These are just a few examples of free artificial intelligence applications that may be useful for public innovation labs. There are many more tools and libraries available, depending on the lab’s specific needs.

Precisely, this invitation to concreteness was also suggested to us by the expert Adriana Freitas , a specialist in impact strategies and exponential businesses, who also added a new reflection: not to assume that AI brings “new things”, but rather that it gives you a “return” (response, result, etc.) to a request that we must know how to limit (format, etc.).

It certainly seemed that “knowing how to ask” or “limit” was something to incorporate into our research. Thus, we focused on solving some of the most “basic” and constant operational needs in labs (and in many other ecosystems): reading, writing, designing and sharing knowledge. Furthermore, we established that these “testable” applications should meet some criteria (preferably):

• A sufficiently useful free version without major bugs.
• Low learning/implementation cost.
• Usable from any browser, without requiring local installation.

Based on these criteria, we browsed the web, app search engines, and the suggestions of chatbots, and tested several applications that already seemed to be gaining traction and had the potential to meet our needs. Thus, from Coboi lab, we shared on social networks our experience with the following applications:

→ Deepl: translates texts, entire documents and entire web pages to the desired language, also offering suggestions and context-based alternatives. Its Chrome extension allows real-time translation of selected web text.

Link: https://www.deepl.com/

→ Mixo: based on a brief written description of an idea, it generates a proposed name, logo, and visuals to give a minimum identity to that idea. It builds a landing page with text and value propositions based on how it interpreted your input.

Link: https://www.mixo.io

→ Canva: a well-known design tool that has already incorporated some AI-based features. By accessing the “apps” menu, and selecting the “text to image” feature, Canva will create an image from the text description you write. The resulting image can be used within any design we are making, and we can apply certain styles (for example, “photography type” or “drawing type”, etc.).

We value that it is a very easy application to understand and use, it is 100% free and no installation is needed on the local computer.

Link: https://www.canva.com/

→ Bearly: it helps summarize many of our readings, such as blog posts, official bulletins, and other online references. It offers multiple functions (e.g., creative generation, grammar correction) but is especially useful for summarizing content. In Chrome, it can summarize the content of the web page you’re viewing. Its desktop version expands functionality with a chat assistant linked to ChatGPT or GPT-4.

Link: https://bearly.ai/

→ DALL.E: generates images from a written description. It is not entirely free, as you need credits to generate your images, but we have found that the Bing image search engine (Microsoft search engine, #Google ‘s “competition” ) allows you to use DALL·E for free.

Link: https://openai.com/dall-e-2

→ Rask: translates a video into another language, while using the original speaker’s voice.

Link: https://es.rask.ai/

At Coboi lab we will continue experimenting with applications based on artificial intelligence as we want to remain active in this innovation and be part of the broader debate about the future and the role of artificial intelligence in it.

Therefore, if you’d like to recommend an application, we would be glad to test it.