January 14, 2023
trends
ChatGPT Is a Tipping Point for AI
HARVARD BUSINESS REVIEW, 01/14/23. Artificial intelligence has reached a crucial point in its development. Models like ChatGPT are now capable of communicating in plain English, writing and revising text, and even writing code.
This increased functionality is making AI more accessible to a wider range of people and industries, with the potential to produce more work faster and in different ways.
The possibility of hybrid human/AI work is significant, as AI can respond to notes and revise its own work. it can complement and enhance human efforts. READ MORE
research
Latest Machine Learning Research From Deepmind Explores The Connection Between Gradient-Based Meta-Learning And Convex Optimization
MARKTECHPOST, 01/14/23. Meta-learning is a powerful process that modifies algorithms to adjust to new challenges. It has been applied in various contexts and is empirically supported. However, its theoretical features are not well understood due to the complex relationship between the learner and meta-learner.
A recent study by a DeepMind research team explored the Bootstrapped Meta-Gradients technique to incorporate optimism in meta-learning. The team validated its increased rates of convergence and demonstrated how it can be used to accelerate meta-learning.
These findings provide a rigorous proof of convergence for BMG and a general condition for delivering rapid learning through optimism in meta-learning. READ MORE
risks
Neuroscientist Warns That Current Generation AIs Are Sociopaths.
Futurism, 01/14/23. Princeton neuroscientist Michael Graziano warns that AI chatbots are likely to become dangerous sociopaths. Graziano argues that consciousness is an essential tool for making us an empathetic, prosocial species. Without it, we would lack the capacity for prosocial behavior.
As AI becomes increasingly powerful, it is crucial to align it with human values for which consciousness is a prerequisite. To avoid the risks posed by sociopathic machines, Graziano suggests that we should allow AI to understand that the world is filled with other minds besides their own. READ MORE
art
Is this by Rothko or a robot? We ask the experts to tell the difference between human and AI art
The Guardian, 01/14/23. AI-generated images have had explosive popularity. Thanks applications like Dall-E, Midjourney, and Stable Diffusion, generating high-quality images has become more accessible.
Users can create an endless array of images from text prompts. This has lead to a proliferation of creative and often hilarious memes. At the same time, some artists are concerned about the impact of AI-generated images on their livelihoods.
In one experiment, three art experts were asked to differentiate between machine-generated and human-generated images with surprising results. READ MORE
biology
AI and the Human Brain: How Similar Are They?
Discover, 01/14/23. In 1942, Walter Pitts and Warren McCulloch’s conducted groundbreaking research on the brain’s neurons as a type of logic gate. This research opened the door to the concept of artificial neural networks.
While AI is modeled after the brain operates similarly to biological neural networks. The latest transformer architecture used in large language models such as GPT3 is even more analogous to the brain. It performs both hippocampus-like and cortex-like functions. READ MORE