Intelligence as a measure of how we treat others (013)
Welcome to Artificial Insights: your weekly review of how to better collaborate with AI – featuring an interview with futurist Lydia Caldana.
Happy last day of July and welcome to Artificial Insights.
In the last few weeks I have been thinking about the implications of being surrounded by multiple types of intelligence. In all likelihood, we as a species will have to contend with losing the top spot in the evolutionary food chain. Whether we ever deserved (or occupied) that spot is arguable, but the advent and discovery of surprising new types of intelligence cohabiting with us should invite the question of what it even means to be a species of great intellect.
If we were to identify or even create species for whom we are nothing but local fauna, what would that entail? Would that force a collective reconsideration of priorities and values, or would we double down on our archaic constructs of scarcity and resource hoarding? Would it invite us to strengthen our collective bonds and promote mutual advancement? Which of these attitudes, if any, would a superior intelligence expect of us?
This week’s newsletter features videos and articles reflecting on this, together with an insightful interview by futurist friend Lydia Caldana. Do reach out if this resonates.
This online panel moderated by Raziye Buse Çetin with participants Mushon Zer-Aviv, Nora Al-Badri and Vulane Mthembu examines the need to decolonize artificial intelligence, emphasizing that AI is underpinned by legacy knowledge systems and power structures. The panelists assert that the historical data used for training AI reflects and continues to reinforce colonial ideologies, particularly noticeable in the dominance of global North visual references in AI datasets. To truly decolonize AI we must critically address infrastructural issues, data concentration, and the hasty pursuit of new AI advancements.
Ilya Sutskever, chief scientist and co-founder of OpenAI, discusses the potential of large neural networks and the power of artificial superintelligence. Sutskever and Sven Strohband explore the idea of merging humans with AI as a possible pathway towards AGI. Deep and insightful.
Stephen Wolfram discusses his work on generative AI space and the mental imagery of alien minds, demonstrating how AI generates images by modifying neural networks, showing the progression from random pictures to more defined images through the capture of regularities in billions of images on the internet Diffusion). Wolfram explores the concept of interconcept space, where feature vectors representing concepts can be visualized as points in a multidimensional space and talks about the "ruliad," representing the entangled limit of all possible computations. Related blog post
Media theorist Douglas Rushkoff delivers a piercing discussion on the damaging repercussions of the "billionaire mindset," focusing on various facets from tech moguls' obsession with apocalypse-proof bunkers to society's relentless pursuit of digitized 'womb-like' comfort. Satirically dubbed as the 'intellectual dominatrix,' Rushkoff throws critical questions to the rich regarding their heavy reliance on technology and economics, and lack of human empathy.
This insightful piece, co-authored by journalist and computer science expert Tim Lee and cognitive scientist Sean Trott, delves into the complex world of LLMs like ChatGPT. The duo aims to demystify the processes that power LLMs and provide the general public a better understanding of these models, without having to grapple with advanced math or technical jargon. The article presents how LLMs use billions of words of everyday language for training, and how they predict the next word in a string. Great primer for anyone interested in the intersection of AI and cognitive science.
The origins of ancient inscriptions are often shrouded in mystery. Writing carved into stone millennia ago can be hard to read and is often missing entire sections of the text. Researchers Yannis Assael and Thea Sommerschield unveil Ithaca - a tool utilizing the power of deep neural networks to restore and spatially-temporally map ancient texts. Ithaca brings a fresh take on interpreting old inscriptions by filling missing information in the text and offering historians a spectrum of hypotheses to evaluate resulting in 72% accuracy in bridging textual gaps by combining AI with human expertise. Read the paper for more insight.
Long reads from Substack
Tackling biases and promoting inclusive technology
This week’s edition features an interview with Lydia Caldana, a friend and foresight strategist focusing on the global south, gender, sustainability and youth. She’s the founder of Future Resources, a platform for womxn and non-binary folks in foresight, strategy and innovation and hosts active WhatsApp communities and database with a wealth futures-related sources of information, companies and people.
What excites and concerns you about AI?
What excited me about AI is that the first stage of most projects (the mapping of what is happening / hypotheses / general data collection) will be sped up without compromising quality. There will have to be another step prior to this in which comprehensive prompts will be inputed - and here is where biases will come into question -, but if done well, this can spare people of this first round of (often) obvious and non-intriguing findings.
What concerns me about AI is what concerns me about all technology: who and how it is being used. I do not wish for AI to be yet another tool developed by and for white Global North people that excludes and discriminates against non-native English speakers, non-white, non-(Westernly)educated, non-cis male, non-able bodies, non-industrialized, non-rich, etc.
What strategies should leaders adopt to embrace more diverse futures?
Leaders should have actual diverse people on leadership positions so that strategic decision-making on how and why these tools are created, all the way down to operational things like features, language nuances, etc. I love Nigerian data scientist Wuraola Oyewusi’s initiative to teach AI concepts in Yoruba language.
How are you using AI yourself today?
I am currently partnering with a scriptwriter to envision a future in which femininity is a valued trait in the labor market. We’re using MidJourney, and the project will come out soon.
Has your work as a futurist made you feel more prepared to deal with the present?
Definitely. By monitoring futures signals and strategizing possible future outcomes, I feel empowered to take action in the present.
Diffusion models are a class of generative models that learn a data distribution by training a neural network as a denoising autoencoder. This process is akin to starting a random walk from a known data point and ending at a point sampled from a simple known distribution. The trained network learns to reverse this random walk, progressively denoising the Gaussian noise and returning the walked path back to the sampled data point. Diffusion models have been found highly effective when used for generating realistic synthetic images, an application that's becoming increasingly critical in deep learning for tasks like image synthesis, inpainting, super-resolution, among others.
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 wanting to learn about AGI.
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.