• Turing Post
  • Posts
  • FOD#6: AI (has been) everywhere (for a long time)

FOD#6: AI (has been) everywhere (for a long time)

Who else jumps on the AI bandwagon, what spatial computers might mean for us, the promises of AlphaDev, and no extinction this week!

Froth on the Daydream (FOD) – our weekly summary of over 150 AI newsletters. We connect the dots and cut through the froth, bringing you a comprehensive picture of the ever-evolving AI landscape. Stay tuned for clarity amidst the surrealism and experimentation.

Today, we discuss the desire to jump on the AI bandwagon and showcase your AI to everyone. We delve deeper into the universe of spatial computers and speculate on its future implications. We celebrate AlphaDev, which discovered a new sorting algorithm. We touch upon alarms about extinction and explore the possible connections with TESCREAL ideology, and also share a few other fascinating reads that you will love.

AI Bandwagon in the froth on the daydream —ar 16:9 —v 5.1 by Midjourney

And remember, some of the articles might be behind the paywall. If you are our paid subscriber, let us know, and we will send you a pdf.

AI (has been) everywhere (for a long time)

"What does it mean that I need to send a fax? Why don't they have email?!" Soon, we may find ourselves utterly frustrated if a company, application, or website doesn't feature a chatbot or some other form of generative AI tool. With breakthroughs in Large Language Models (LLMs) such as ChatGPT, which excel in analyzing, reasoning, and conversing, as well as generating text, images, and videos from text prompts, generative AI tools represent a paradigm shift in how we interact with technology and each other. These tools are becoming increasingly integral to our everyday lives.

Following in the footsteps of Microsoft with their Co-pilot, Snapchat with Snap AI, Adobe with Firefly AI Image Generator, and Salesforce with AI Cloud – all of which have launched various features integrating generative AI into their core businesses – Mark Zuckerberg plans to integrate text, image, and video generators across all platforms, including Facebook, What’sApp and Instagram. (You can listen to Mark Zuckerberg discussing the Future of AI at Meta with Lex Fridman here.)

So many AI news! As if all these companies never used AI&ML before. They have, in fact, been employing Data Mining and Analytics to extract insights from large-scale user data for years, improving user experiences and talent management. They've used Reinforcement Learning to train AI agents in order to optimize engagement and user experiences in areas like content ranking and ad delivery, and employed Deep Learning to utilize neural networks with multiple layers to train models for a range of tasks, including image recognition, speech recognition, and language translation - among many, many other AI and ML techniques. But now, they find it necessary to be more vocal about it. Some thinks it’s a sign of a bubble.

Governments are quick to jump on the AI bandwagon too

Governments are quick to jump on the AI bandwagon. As Congress develops legislative proposals and educates itself through bipartisan briefings to keep pace with this rapidly advancing technology, Microsoft is reportedly bringing OpenAI’s GPT-4 AI model to US government agencies. Neuron newsletter believes this could be the collaboration of the century. Government agencies handle and process enormous volumes of information, and the potential applications of AI are vast:

  • Sorting through mountains of paperwork with ease.

  • Providing 24/7 customer support.

  • Summarizing field reports in seconds.

Kudos to both us and them if this indeed comes to fruition! They could already learn from the city government of Yokosuka, Japan, which reportedly uses ChatGPT to increase work efficiency by summarizing records and documents.

MetaApple

Apple was actually the first company to make me realize the general public's lack of AI&ML knowledge in 2019. If you think about it, a regular iPhone is stuffed with machine learning techniques. It utilizes it in features such as Siri, FaceID, photo categorization, Animoji and Memoji, predictive typing, and personalized recommendations, to name a few. These functions are enabled by the Apple Neural Engine (ANE), a component dedicated to machine learning computations, which started with the A11 Bionic chip. Moreover, Apple's Core ML framework allows developers to integrate machine learning models into their apps, supporting a variety of model types, including neural networks, tree ensembles, and support vector machines.

This time, Apple also didn't jump on the "I also use AI!" bandwagon. They didn't mention AI at the annual WWDC, but did finally bring attention to machine learning. By downplaying the role of AI and letting it quietly enhance our experiences without emphasizing its mechanics, Apple's strategy worked well with iPhones. Users could enjoy the fruits of this technology without worrying about potential existential threats AI might pose. This is a deliberate position: showcase technology, captivate everyone, but don't delve too deeply into what it actually is and how it functions.

However, with the introduction of Vision Pro – Apple’s spatial computer with eye-tracking technology, it becomes essential to understand what this technology might mean for us. VisionPro is based on visionOS which has a new real-time execution engine: “that guarantees performance-critical workloads, a dynamically foveated rendering pipeline that delivers maximum image quality to exactly where your eyes are looking for every single frame”. According to Stratechery, real-time operating systems are crucial for applications that demand high reliability and performance, much like a vehicle's operational software.

Does this indicate that Apple is venturing into self-driving cars? We can't be certain. But with Vision Pro, they've opened the door to the possibility of the metaverse. Semafor writes, “Mark Zuckerberg reoriented his company around the concept, but his vision was met with skepticism from investors and the media. A lot of that backlash comes from Facebook’s bruised reputation, which stands in stark contrast to Apple’s teflon luster. The tech press, which received 15 to 30-minute guided demos of Apple’s Vision Pro headset on Monday, has been gushing about the device. Apple is taking a gamble that it can reverse the ho hum consumer response to these products. Meta had already made that gamble and until Monday, that was a very lonely place to be.

Although we will have to wait to try Vision Pro in real life, if you're eager for a sneak peek into the Vision Pro test experience, here are a few highlights:

With all this hype, it's funny to see The New York Times trying to jump in with their two golden prompts. Somebody needs to tell them that there's a whole prompt industry out there already. Or, at least, someone could send them the GPT best practices that OpenAI recently put together on their website. By the way, Harvard Business Review argues that AI prompt engineering is not at all the future. What do you think?

Wow, AlphaDev!

Last week, Sam Altman encouraged South Korea to supply chips in AI boom. But what else can we do when the scarcity of GPUs is so apparent? Here comes AlphaDev! Its importance lies in its potential to revolutionize the optimization of the world's code through AI. By iterating on DeepMind's previous AlphaZero and MuZero models, AlphaDev uses reinforcement learning to discover faster and more efficient algorithms, with remarkable results already seen in the sorting and cryptography domains. The enhancements by AlphaDev extend the boundaries of traditional algorithmic development, offering up to 70% speed enhancements on key algorithms. This significant advancement is particularly valuable as AI models begin to strain the physical limits of computer chips, necessitating more efficient computation methods.

AlphaDev's contributions also signify a new age in foundational algorithm research, as it challenges the limits of human-led efforts by achieving superior results in areas untouched for decades. Its improvements have already been integrated into two C++ libraries, demonstrating the real-world impact and adaptability of these AI-discovered solutions. Moreover, it opens doors for AI to tackle optimization challenges in other domains and for larger, more complex algorithms, representing a transformative step towards fully AI-optimized coding. The take from The Batch is that “deep learning’s potential for synergy between humans and machines: People supply an algorithm (such as matrix multiplication) and AI accelerates its runtime.” A truly promising era of human-machine cooperation!

But let’s get back to the risk of extinction. Are we there yet?

Not this week. I’d say this week we are still doing pretty well. Some even think we might be saved soon. Last week's narratives highlighted the complexity of AI discourse, its risks and rewards, and the importance of collaboration, diversity, and judicious regulation to ensure a beneficial trajectory for AI.

While Sam Altman advocates for collaboration between the United States and China to address the risks associated with AI development, and Elon Musk engages in discussions with the Chinese government to establish rules for using AI, AI Snake Oil criticizes the concept of AI non-proliferation via licensing, deeming it infeasible and likely to concentrate power in the hands of a few big tech companies. Instead, it suggests a diverse group of entities, including academics and NGOs, for evaluating and developing AI models to address risks. The article in the The Road To AI We Can Trust newsletter warns that the wrong regulation could lead to adverse outcomes, “If big tech writes the rules, without outside input, we are unlikely to wind up with the right rules.”

The concern over an 'AI savior complex' was presented in an interview with Kyunghyun Cho, an AI researcher and associate professor at New York University. Cho suggests a misplaced emphasis on AGI and existential risk by influential entities in Silicon Valley. "I’m not a fan of Effective Altruism (EA) in general. And I am very aware that the EA movement is driving the discourse around AGI and existential risk. I think there are too many people in Silicon Valley with this kind of savior complex. They all want to save us from an inevitable doom that only they see and they believe only they can solve." Effective Altruism is part of the TESCREAL ideology (an acronym for Transhumanism, Extropianism, Singularitarianism, Cosmism, Rationalism, Effective Altruism, and Longtermism) that is allegedly influencing the thoughts and actions of the AI tech elite. There's a lot to discuss here, but we'll table that discussion for better times (email us if you want to write a column about it).

So who does think that we might be saved soon by AI? It’s Marc Andreessen, one of the most influential investors. He extensively explained in his post "Why AI Will Save The World". He argues that AI is an enhancer rather than a threat, likening its impact to that of electricity and microchips. He believes AI will augment human intellect, creating more opportunities, jobs, and industries. He also cautions about risks like regulatory capture and the international AI race, particularly with China. Overall, it’s a very bullish, investor-oriented blog. But being pro-business doesn't mean it’s wrong. Andreessen is renowned for his pro-business blog posts that call to action and are generally accurate, with the main motto “Let American firms be allowed to build as fast and as aggressively as they can, you can also check his popular posts "Why Software Is Eating the World" (2011) and "It's Time to Build" (2020).

To maintain your optimism, check out pessimistsarchive.org (recommended by Andrew Ng). The site describes fears of non-fiction novels corrupting youth, elevators causing brain fever, cars (termed "the devil wagon") on a mission to destroy the world, and recorded sound harming babies.

If you want to dream about a peaceful world, take a look at Zuzalu, a pop-up city community in Montenegro.

What we also read/paid attention to last week:

Some practical stuff:

Just to read:

We hope you liked it. Please feel free to share with everyone 🤍

Join the conversation

or to participate.