AI is a toolset, not a replacement. Programmers are still needed! - DSBoost #27

💬 Interview of the week

This week we interviewed Martin Kosobud, who is a Fintech account manager for Mastercard. Enjoy:

• What did you study/are you studying (if your background is different from DS, how did you end up in the field)?

  I studied economics and finance in the Netherlands and subsequently in the UK. However, more recently, I have been pursuing a Data Science master's at the University of Edinburgh. My journey into Data Science began primarily during my tenure at Amazon and was further developed at my current employer, Mastercard.

• What are your favorite resource sites and books (ML/AI)?

  I love Harvard’s CS50, the YouTuber Fireship, and a few others. However, lately, I've been obtaining most of my information about recent progress directly from training materials, articles, and meetups created by Microsoft, Google, and OpenAI. Of course, the university plays a huge role as well. One book that significantly influenced my thinking about data visualization is "Storytelling with Data" by Cole Nussbaumer Knaflic, and I'm glad that the insights from this book are now widely recognized in the business world.

• What got you into your current role (portfolio, certification, etc.)?

  Currently, I am a UK Fintech account manager for Mastercard. Although I could have pursued a different path in Data Science, I decided to gain a slightly different experience at work, given that I am already studying DS at university. Before this role, I was part of an Insights and Optimization team.

• What do you enjoy the most in your work?

  I find the combination of consumer contact and the possibility to enrich conversations with data particularly enjoyable. Having a data-driven relationship makes the conversation more engaging and interesting!

• What tools do you use the most/favorite tools?

  For programming, I use Python most of the time, and occasionally a bit of R. I enjoy working within JupyterLab. As for packages, it's quite standard: Pandas, NumPy, Scikit, Seaborn, and similar. Lastly, I prefer using Tableau for data visualization. All of these tools, along with the recent GPT-4, have large communities supporting them, making development quick and relatively easy.

• Do you use ChatGPT or other AI tools during your work? If so, how do they help you? Do they change your approach to problems?

  Yes! While I must adhere to confidentiality constraints, I often use AI tools like ChatGPT for improving wording or finding useful code snippets. ChatGPT's browse search functionality was recently removed, which is a shame, but I hope it will be back soon. These AI tools have significantly changed my workflow, as they allow me to accomplish many tasks faster and provide useful boilerplate to guide me in the right direction.

• What is your favorite topic within the field?

  Time series analysis is closest to my heart, although I would love to become proficient in Neural Networks and Reinforcement Learning. Time series analysis is something I frequently work on, specifically focusing on seasonality analysis and trend analysis. In my previous role in insights and optimization, we often tracked the performance and seasonality of restaurants, for example.

• Which one of the recent AI/ML models will have the most significant impact on the industry in your opinion?

  I believe reinforcement learning models like AlphaGo and its successors from DeepMind (AlphaZero, MuZero) have made waves in the industry by solving complex games that require long-term strategic planning. Of course, more recently, models like GPT-3 and GPT-4 have kickstarted the current huge AI craze.

• What are you currently learning or improving (topics you are interested in nowadays)?

  Recently, I've become more interested in computer vision since my fiancée is setting up a dental clinic, and there can be some potential uses for it. One of the recent advances I've heard about is the use of Deep learning models to detect dental caries (tooth decay) in its early stages from dental X-rays or scans, potentially even before they're visible to the human eye. While I may not develop something similar myself, this technology is currently live and quite impressive!

• What is the biggest mistake you've made? (preferably DS related)

  In the field of Data Science, attention to detail and double-checking are crucial. I must confess that a few times I accidentally switched column X with column Y and similar mistakes, and sometimes it was tough to admit this to my manager. However, I always did, and fortunately, I managed to avoid any potential problems. From these experiences, I've learned to be more precise and structured in my work.

• What is your most significant achievement? (preferably DS related)

  One of my most significant achievements was a few years ago when I was working at Amazon. I developed an Excel model that allowed for verifiable savings of hundreds of thousands of dollars through efficiency gains. That was quite a highlight!

• Can you share a fun fact about yourself?

  In my free time, I enjoy designing and creating jewelry!

🎙️ Podcast of the week

Why AI is Eating the World with Daniel Jeffries, Managing Director at AI Infrastructure Alliance

Key takeaways:

• The future of human-computer interaction lies in AI becoming the central interface for all our software interactions. This shift envisions a world where we can communicate with AI in a conversational manner, much like interacting with a knowledgeable friend or colleague.

• Open source plays a vital role in AI's advancement. By enabling developers to experiment and build upon existing models, it fosters innovation and improves outcomes.

• AI faced some historical challenges called AI winters. These periods were characterized by limited success, leading some to doubt the potential of AI. One crucial development was the release of Stable Diffusion as an open-source solution, which marked a turning point.

• AI is now integrated into all software, changing how we work with it. 'Ambient AI' is rising and will be everywhere, from supply chains to doctor's visits.

Note: Ambient AI refers to the integration of artificial intelligence seamlessly into our daily lives and environments, making it a natural part of our surroundings and experiences. Unlike traditional AI interactions that require explicit commands or user inputs, Ambient AI operates in the background, understanding context and proactively providing assistance or information without requiring direct prompts.

Embrace AI as a powerful toolset, not a replacement. Brilliant programmers are still needed. Artists, don't worry, your creativity remains essential. Adapt, level up your skills, and enjoy the amazing possibilities AI brings to every industry. This is an exciting time to be alive, so embrace it with love and enthusiasm.

🧵 Featured threads

🤖 What happened this week?

• OpenAI has filed a trademark application for 'GPT-5', and the developer Siqi Chen anticipates completing its training by year-end, potentially leading to AGI capabilities. They also argued that Code Interpreter is actually GPT-4.5!

• Llama 2, a previously smaller 7B model, now boasts an impressive 32K context window, enabling the AI to process longer documents and hold more extensive conversations without losing important information. This significant development comes just one week after the model's initial release. You can try it in together.ai.