'AI Has The Potential To Be More Transformative Than the Rise of the Internet'

What's your job and how does it involve AI?

As the Head of Data and Innovation, I lead a team that's passionate about leveraging AI to transform the educational landscape. At Docsity, we've curated a vast collection of user-generated study documents, and with our tailor-made AI services, we convert them into easily digestible summaries, interactive chatbots, dynamic mind maps, and engaging quizzes. By harnessing the power of AI, our aim is creating a more accessible, personalized, and efficient learning environment for students everywhere.

Do you believe AI will bring a bigger technological revolution than the internet?

I believe blindly that AI has the potential to bring about a technological revolution that could be extremely more transformative than the rise of the internet.

The internet has undoubtedly reshaped the way we communicate, access information, and conduct business. However, AI's ability to analyze, learn, and adapt to complex data and environments could unlock new levels of innovation and efficiency across various sectors, from healthcare and education to transportation and energy management.

Is AI more of a chance or a challenge for humanity? Why?

Past signs that every turning point in human history has involved navigating the complex interplay between opportunities and challenges. The rise of AI is no different. As we embrace the transformative potential of it, we must also address the ethical, social, and economic implications it presents.

This includes fostering responsible development, promoting transparency and fairness, and ensuring that the benefits of AI are equitably distributed. Ultimately, the key to unlocking AI's full potential lies in our ability to adapt and evolve alongside the technology, learning from the lessons of the past and working together to create a future where AI serves as a powerful tool for human progress and well-being.

Which Ai tool do you personally use the most?

As part of my daily routine, I actively engage with a variety of AI tools, including ChatGPT, Mixtral Large, and Claude, to explore new possibilities and enhance my productivity. These AI tools serve multiple purposes, such as testing APIs, experimenting with cutting-edge algorithms, generating the draft for high-quality business documents, and conducting research on emerging domains. Moreover, I leverage the capabilities of these AI tools to perform prompt engineering, which allows the team to create more targeted and effective queries, ultimately leading to better results.

Do you prefer: ChatGPT or Gemini?

Mix of well known flagship models like ChatGPT, Claude, Gemini and open source models, as well as open-source models such as Mistral, Mixtral, Llama3 or even the infinitesimal Phi-3 size.
Each AI model brings unique capabilities and perspectives to the table, and by combining them semantically based on specific requirements, I can achieve more accurate and contextually relevant results. In essence, I believe that the true power of AI lies in its diversity and the ability to combine different models to create a more comprehensive and adaptable AI solution.

Do you trust Gen AI benchmarks?

While AI benchmarks can provide valuable insights into the performance of various models, they may sometimes offer a limited perspective on AI capabilities when not accompanied by the right data, prompt engineering, or even the need of business related fine-tuning. Relying solely on these benchmarks can lead to an incomplete understanding of AI performance in real-world scenarios.

Comparing AI models using only benchmarks is similar to evaluating a person's learning abilities exclusively through multiple-choice questions, which may not accurately reflect their problem-solving skills in more open-ended situations. Instead, I believe in a more comprehensive approach that combines A/B testing and feature flag releases to assess AI performance in real-world business and end-user cases. By testing different LLM providers, model sizes, and context lengths, we can gain a more accurate understanding of AI capabilities and make informed decisions about their implementation. This approach ensures that we're evaluating models in a more holistic and practical manner, ultimately leading to better outcomes and more effective innovative solutions.