You must learn to proceed without certainty (023)
Exploring many kinds of human-machine augmentation.
Happy Monday and welcome to another edition of Artificial Insights, where we try to make sense of the torrent of knowledge being created around AI and augmented creativity. This week’s edition features your usual cross-section of interviews, lectures and insights into what I believe to be shaping the near future of human-machine interactions.
Last week was an amazing opportunity to connect with readers online and in real life. Lecturing about Centaurs in London and Amsterdam the past few weeks was a blast – the metaphor resonates and participants surprised me with insights taking the concept even further. We are preparing an Envisioning-themed and annotated deck of the trends and technologies presented, which we plan to share with you all in the coming weeks.
Question: would something like a WhatsApp group to share links and meet other readers be valueable to you?
Denis Shiryaev on Twitter figured out how to circumvent Bing’s CAPTCHA filter using visual prompt engineering.
What if your text-to-image prompts supported rich text like different fonts, sizes and colors?
Mira Murati's Vision 🎨
Interview with OpenAI CTO Mira Murati on her pivotal role in molding generative AI through her stewardship of ChatGPT and DALL-E, shedding light on her journey from Albania to OpenAI.
Underscores Murati's instrumental role in propelling OpenAI’s products, emphasizing her unique journey and vision towards achieving AGI, offering a deep dive into the persona often working behind the limelight.
Highlights the critical alliance between OpenAI and Microsoft, mediated by Murati, hinting at a broader collaborative landscape in AI, crucial for propelling OpenAI towards its ambitious goal of AGI.
Brings to fore Murati’s balanced outlook on AI’s competitive landscape, her proactive approach towards addressing copyright issues, and the necessity of regulatory measures, offering a grounded perspective amidst the AI euphoria.
Accelerating AGI 🖇️
Software development legend John Carmack and Dr. Richard Sutton have formed a partnership aiming to expedite the development of Artificial General Intelligence (AGI).
The collaboration marks a significant step towards fostering AGI development, leveraging Carmack's and Sutton's notable expertise.
This partnership is a strategic move, combining reinforcement learning acumen with software engineering prowess to tackle AGI's complex challenges.
Their ambitious goal of developing a genuine AI prototype by 2030 sets a tangible target in the evolving AGI landscape.
Economic transformations through AI 📈
Paul Krugman explores the potential impact of AI on the economy, drawing parallels between AI and past technological booms. The speculation centers around AI's capacity to augment productivity, akin to the IT-driven boom of 1995-2005. Goldman Sachs estimates a 15% productivity surge over a decade, with significant effects on high-end administrative jobs, possibly reducing income inequality, a deviation from past technological advancements.
Highlights AI's potential to significantly boost productivity, drawing a parallel with the IT boom of 1995-2005, offering a historical lens to gauge AI’s prospective economic impact.
Goldman Sachs' projection of a 15% productivity increase over a decade, underlining the substantial economic transformation AI could drive, even without achieving 'true' intelligence.
Discussion on income inequality dynamics, suggesting a unique scenario where AI might lower income inequality, contrasting with common technological progression that often exacerbates it.
Exploring skill emergence in scaled-up language models 🦜
Sanjiv Aurora delves into the realm of emergence in language models (LLMs), illustrating how scaling up LLMs augments performance on benchmark tasks significantly. Through plots, he elucidates the performance escalation and how minor alterations in scaling laws can substantially affect the model's overall behavior. Although not establishing AGI's existence, he underscores the potential for complex skill emergence in LLMs.
Through visual plots, Aurora provides a compelling depiction of how scaling enhancements lead to performance upticks in benchmark tasks, elucidating the impact of minor scaling law adjustments.
By showcasing the model's capability to generate constraint-adhering responses, he highlights a notable advancement in LLMs, pushing the boundaries of what these models can achieve.
Although not a confirmation of AGI, the discussion opens a window into the substantial potential of complex skill emergence in LLMs, marking a significant stride towards understanding and harnessing the power of large language models.
AI’s Potential Knowledge Monopoly 📚
New Yorker article exploring the evolution from human-generated knowledge on platforms like Stack Overflow to AI systems like ChatGPT and Bard ingesting and monopolizing web knowledge. Jeff Atwood, Joel Spolsky, and Sam Altman are cited concerning the shift from crowdsourced web knowledge to AI-driven knowledge synthesis, highlighting the decline in human contributions to online repositories and the potential for AI to self-generate knowledge through synthetic data.
Reveals a potential shift from public, human-generated knowledge to privatized, AI-synthesized knowledge, threatening the ethos of platforms like Stack Overflow.
Discusses synthetic data's potential in augmenting AI's self-learning, contrasting it with the current dependency on human-generated data for training AI models.
Raises concerns about the concentration of knowledge in AI and the potential dangers, juxtaposed against the convenience and efficiency of AI-driven knowledge synthesis.
Long reads from Substack
A fundamental supervised learning algorithm which aims to establish a relationship between a dependent variable and one or more independent variables by analyzing a given dataset. The goal of regression is to predict numerical continuous values for the dependent variable based on the values of the independent variables. In regression, the dependent variable is often referred to as the target variable, while the independent variables are known as features or predictors.
If Artificial Insights makes sense to you, please help us out by:
Subscribing to the weekly newsletter on Substack.
Following the weekly newsletter on LinkedIn.
Forwarding this issue to colleagues and friends.
Sharing the newsletter on your socials.
Commenting with your favorite talks and thinkers.
Artificial Insights is written by Michell Zappa, CEO and founder of Envisioning, a technology research institute.
You are receiving this newsletter because you signed up on envisioning.io.