The Last 10 AI "Breakthroughs" that are Sinking Up the Data Landscape
Navigating AI Growth While Maintaining Lean Data Practices Is Hard.
Last week's AI breakthroughs (or let’s say announcements as most of them are not in production yet) have transformed the AI landscape, presenting both exciting opportunities and challenges for Lean Data professionals. As we delve into these developments, we must recognize the significance of stability, comfort, and familiarity in driving successful adoption. People need a sense of stability and reliability to navigate this new way of working with machines. By prioritizing these factors, we can shape the future of Lean Data and ensure that individuals feel confident and comfortable embracing AI technology. Let's explore the implications of these breakthroughs and how they will impact our work with machines.
1. Multimodal ChatGPT: A New Era of Interaction
OpenAI's ChatGPT has finally disclosed its vision of a multimodal model, now boasting voice and image capabilities. While voice and image features offer opportunities for intuitive data exploration, concerns should arise regarding accountability and decision-making in business settings.
The role of AI in interpreting results and the need for human oversight must be considered. Building lean data workflows requires careful evaluation of efficiency, accuracy, and potential risks. Striking a balance between AI advancements and human involvement is crucial.
For example, we could use voice interactions to ask specific questions about data insights and leverage image capabilities to visually analyze complex patterns or anomalies, but we still need a human in the loop to validate that. The possibilities are intriguing, and I'm excited to see how these features evolve and unlock new opportunities for lean data professionals.
→ Learn more about these new features
2. DALL.E 3: Simplifying Text-to-Image Transformation
OpenAI's latest text-to-image model, DALL.E 3, has made significant strides compared to its predecessor, eliminating the need for complex prompt engineering. While this model brings significant advancements and eliminates the need for complex prompt engineering, it also introduces potential downsides. One such downside is the ability to include text in images, which may lead to the proliferation of misleading or inappropriate content, particularly in the form of memes.
This could pose challenges for professionals who rely on visualizing data insights and trends, as the accuracy and reliability of generated images may be compromised. It is important to remain vigilant and critically evaluate the output of AI models to ensure the integrity of data visualization in the face of these potential downsides. DALL·E 3 is built natively on ChatGPT, but not yet available. They are still researching “the best ways to help people identify when an image was generated with AI”.
3. Cohere's Chat API: Powering Up Product Experiences
The launch of Cohere's Chat API public beta with Retrieval-Augmented Generation (RAG) allows developers to integrate user inputs, various data sources, and model generations into powerful product experiences with grounded and verifiable responses. RAG is a great piece of tech that incorporates external data sources to improve relevance, accuracy, and timeliness. The Chat API enables a conversational experience, and developers have control over model selection, temperature adjustment, and chat history utilization. Citations can also be included to enhance trust and credibility. This breakthrough could revolutionize how data professionals interact with their products and users, enabling more personalized and efficient experiences.
It’s an approach that we follow very closely at Naas.
4. ChatGPT Browsing: The Non-News
OpenAI's reintroduction of the ChatGPT Browsing feature means that ChatGPT can now re-browse the internet to provide current information with direct links to sources. This feature is a great addition since the model was trained on data before 2021 but the reliance on web browsing introduces the challenge of verifying the credibility and trustworthiness of the sources. It is crucial for users of the ChatGPT Browsing feature to exercise critical thinking and fact-checking to ensure the accuracy and reliability of the information obtained.
5. Microsoft's Copilot Integration: Context-Aware Assistance with Potential Downsides
Microsoft's integration of Copilot into its popular products offers the promise of context-aware assistance, revolutionizing how users interact with technology. For Lean Data professionals, this means AI-powered support in data analysis and decision-making, leading to increased productivity and efficiency. However, it's important to consider potential downsides, such as privacy concerns related to data collection and the need to balance relying on AI assistance and preserving critical thinking skills. By addressing these challenges, Microsoft can ensure that Copilot delivers its full potential while upholding user privacy and fostering responsible AI usage. Or, Copilot will become the new Clippy.
6. ChatGPT for GitHub: Streamlining Development with AI
GitHub's introduction of the ChatGPT feature in Copilot lets you talk to your Git repo, streamlining the development process using AI. On paper, you can ask questions, receive guidance on coding best practices, perform code analysis, receive security recommendations, and troubleshoot coding issues without the need to switch contexts. I’ve been using it in VScode and it falls a bit short when it comes to context awareness in my opinion. The chatbot is currently free for individual GitHub Copilot users, but it remains unclear if it will be priced separately in the future.
→ Learn more on Visual Studio Magazine
7. Meta's AI Innovations: Personalizing User Experience
Meta's introduction of a chat assistant with multiple personalities, AI-generated stickers, and AI editing in Instagram, along with the launch of "AI Studio" and collaboration with Ray-Ban on smart glasses, indicates a shift towards a more personalized user experience. If you add on top of this the fact that Meta focuses on open-sourcing various aspects of their work, they are clearly positioned as the leading company in AI. This trend suggests that sharing research, models, and datasets publicly in open source actually enhances competitiveness, especially in the field of AI. Contrary to common belief, open sourcing can be a strategic advantage rather than a disadvantage. I believe that Meta could win business data professionals as they can enhance user engagement and satisfaction with their approach.
8. YouTube's Generative AI Features: Empowering Creativity
Finally! YouTube has announced generative AI features including AI-generated photo and video backgrounds, AI music recommendations, and voice cloning in other languages, aiming at making content creators more creative and productive. This is inspiring as most data professionals must explore new and innovative ways of presenting data insights.
9. Mistral AI's LLM: The Next Open Source Rockstar?
Mistral AI's unveiling of its inaugural LLM, with its 7B model surpassing the Llama-2 13B in all benchmarks, signals a shift towards open source. This could provide data professionals with more diverse options, greater transparency, and increased competition over which model to use in business settings.
I’m currently investigating how we can test Mistral in Naas Chat. Stay tuned.
10. Google's Bard Extensions: A Futuristic User Experience
Google's release of Bard Extensions, integrating the most popular Google apps with Bard is going to be a fantastic user experience. It could revolutionize how Data professionals access and analyze data across multiple apps. Especially considering the need for integration that AI Chat requires if we want to use them in business workflows.
I don’t know what to think about this list, it’s ever-increasing and it's just getting started. These breakthroughs are not just reshaping the AI landscape but are also setting the stage for a future where Lean Data professionals can leverage AI to drive productivity, innovation, and growth, but this will come at a cost.
“AI does not automate thinking.” like Cassie Kozyrkov rightly said, I would highly recommend you to watch it if you have made it this far.
I would also recommend Shubham Saboo’s content for more in-dept updates on the different AI technologies out there. He is the creator of the “Unwind AI” newsletter and created this mindmap as a a summary of this last week 10 breakthroughs:
Thanks for reading, let’s continue to explore this exciting transformation of tech, business and society with data & AI.
We are still actively taking feedback on Naas Workspace, currently developing a strong bridge between the Lab (notebooks) and the Chat. You can stil get your invite on this link: https://bit.ly/3CFSumN
Cheers!
J.