Sentiment Analysis

AI Investment and the Internet Analogy

It's the late 1990s, and there was a gold rush happening. But instead of pickaxes and river pans, people were scrambling for modems and domain names, like Business.com (originally bought for $7.5 million but sold for upwards of $340 million in 2007).

It felt familiar to see that Scale.AI was valued at $13 billion. Unlike Databricks, a cloud software company, Scale.AI isn't the same. Scale.AI expects a 53% profit margin this year, much lower than the 76% average for cloud software companies and 78% for Databricks. Although Scale.AI creates software tools, it mainly provides human workers for AI training, which costs a lot. This is not a bad business, but I don’t think they’ll get to the revenue and margins to make the valuation a reality.

Now, back to the Internet in the mid-1990s. Dial-up internet service providers like the one I worked for struggled to get bandwidth and data center space for modems. People were scrambling to get modems from U.S. Robotics, which boasted 56K kilobit downloads. Modern speeds are about 1,000 to 10,000 times faster than 56K modems.

In 1995, Netscape Communications went public on August 1, and its share prices increased by 108% on its first day. Also in 1995, an engineer at Netscape, Brendan Eich, was given an ultimatum: create a working prototype of a scripting language in just one week. Eich had to design and implement a language that could manipulate the elements of a web page and provide interactive functionalities. It was an ambitious goal, considering the complexity of designing a programming language from scratch. Today, that language is the world’s most popular programming language —javascript.

Then there was Global Crossing, which spent billions laying underwater fiber optic cables, connecting continents at unprecedented speeds. In 1999, during the dot-com bubble, the company was valued at $47 billion, but it never had a profitable year. In 2002, the company filed for one of the largest bankruptcies in history, and its executives were accused of covering up an accounting scandal. On October 3, 2011, Global Crossing was acquired by Level 3 Communications for $3 billion. When they went bankrupt in 2002, those cables didn't disappear. They were bought up for pennies on the dollar and became part of the foundation for our modern, high-speed Internet.

Today's tech giants are doing something similar but on an even grander scale. AWS, Google Cloud, and Microsoft Azure are the new railroad barons, except instead of tracks, they're laying down a web of AI-capable data centers across the globe and producing foundation models. They aren’t scrambling for modems but rather Nvidia’s highly capable GPUs used for AI training and inference.

Also, today’s data centers are different. They are being designed to handle the immense computational power needed for AI. And just like the specialized equipment developed during the Internet gold rush, we're seeing a new generation of tools being created for AI.

Today’s pickaxes and sluice boxes are AI-specific chips. NVIDIA, Intel, and AMD are crafting GPUs and AI accelerators to crunch numbers faster than ever. And just as the gold rush had its innovators who created new mining techniques, we have startups like Groq and Cerebras Systems pushing the boundaries of chip design.

But here's where it gets really exciting. Remember how the Internet bubble helped build out the infrastructure for the digital age? I am writing this in a Chrome browser, an evolution from the Netscape browser of the 1990s. There are innovations that we benefit from even when the companies disappear.

Today, the rollout of 5G networks isn't just about faster phone downloads - it's about creating the low-latency environment necessary for real-time AI applications. Self-driving cars communicate with smart cities in real-time, all powered by this new infrastructure. Just like during the Internet bubble, we are seeing immediate impacts, though nothing like the 24 years after the bubble burst, infrastructure, networks, and even software that set the stage for AI.

Just like the Internet boom created a new breed of tech-savvy workers, the AI revolution is fostering its own wave of high-tech jobs and entrepreneurship. It's as if we're watching a new Silicon Valley spring up, but this time, it's global and all about AI.

Here's the kicker: while some of the exuberance around AI might seem reminiscent of the dot-com bubble, the foundation being built is rock solid. Unlike some of the more speculative Internet companies of the late '90s, the infrastructure being developed for AI has immediate, practical applications.

So, while we might see some AI startups flame out spectacularly (just as we saw with Pets.com or Webvan), the underlying infrastructure—the data centers, the specialized chips, the enhanced networks—is here to stay. And it's setting the stage for a future where AI is as fundamental to our lives as the Internet is today. In essence, we're not just witnessing another tech bubble.

We're watching the birth of a new technological era, one that's being built on a foundation far more solid than anything we saw during the Internet boom. And that's what makes this AI revolution so exciting and transformative.

Brought to you in partnership with

I use Beehiiv to publish the Artificially Intelligent Enterprise. I think it’s the most robust newsletter platform. I am not only a user but an investor.

SAI TL;DR - Lastest AI News for Business Users

• AI Ready: A Free Book from an IT Practitioner - My friend Tony Fatouros is the Director of IT at Mattel (yep, the people who make Barbie). He leads a high-performing change and program management team and shares his experience in his free book, which is a good read.

• AI-Driven Behavior Change Could Transform Health Care - Every aspect of our health is deeply influenced by the five foundational daily behaviors of sleep, food, movement, stress management, and social connection. AI, using the power of hyper-personalization, can significantly improve these behaviors.

• Microsoft and Apple ditch OpenAI board seats amid regulatory scrutiny. Microsoft has dropped its seat as an observer on the board of OpenAI less than eight months after securing the non-voting seat. Apple was reportedly planning to join OpenAI’s nonprofit board, but now It will no longer do so. Is this unnecessary, or is it because regulators are making collaboration too risky?

• OpenAI’s China Ban Doesn’t Apply to Microsoft’s Azure China - OpenAI last month announced plans to crack down on China-based artificial intelligence developers covertly accessing the startup’s technology, which it doesn’t offer in China. However, OpenAI’s conversational AI models are available to Chinese businesses if they sign up for Microsoft’s Azure cloud service, which operates in China through a local joint venture.

• A16Z Stashing GPUs To Win AI Deals- Venture capital firm Andreessen Horowitz has purchased and stashed thousands of GPUs, including many of Nvidia's H100 processors, to win deals for AI startups and further exacerbate the GPU shortage.

• AI's Energy Demands Are Out of Control. Welcome to the Internet's Hyper-Consumption Era. Generative artificial intelligence tools, now part of the everyday user experience online, are stressing local power grids and causing mass water evaporation.

Brought to you in partnership with

I use Synthesia to take presentations and make them into explainer videos using AI-generated avatars. It’s easier to update the videos using text to video than refilming when something changes.

AI in Action - Your Weekly Deep Dive to Mastering Enterprise AI

Sustainable AI

This week, in the AI TL;DR, I chose several news stories that I thought were important because of their relevance to one topic — the need for industry collaboration. I feel like this is needed to create a responsible, safe, transparent AI industry that is sustainable and gets the most benefit from this massive investment.

Given artificial intelligence's growing influence and complexity, there is a clear need for a similar authoritative body. This organization would develop and promote standards for AI to ensure ethical practices, interoperability, and robustness across various platforms and industries, enhancing trust and facilitating innovation in AI technologies.

Industry Collaboration

Apple and Microsoft's recent decision to step away from the OpenAI board underscores the need for a broader, more inclusive approach to AI governance that allows industry leaders to collaborate to prevent collusion and drive collaboration safely. This move highlights the importance of creating an independent organization dedicated solely to AI, free from corporate influences that could stifle innovation and transparency.

Effective collaboration across industry players is vital for sustainable AI development. The IEEE, a leading global organization for advancing technology, exemplifies the power of industry-wide collaboration. The IEEE has been instrumental in fostering innovation and interoperability across diverse technological fields by setting standards and facilitating cross-domain cooperation.

The Linux Foundation is another stellar example of successful collaboration. It has brought together tech giants, including Google, IBM, and Intel, to work on open-source projects that have revolutionized the industry. By adopting a similar model, the AI community can ensure that developments are not siloed but shared for the collective benefit.

I think we need an international consortium that allows stakeholders to collaborate.

• Encourage and enforce interoperability standards to prevent monopolistic practices and encourage a competitive and diverse marketplace. For example, similar to how the World Wide Web Consortium (W3C) manages web standards to ensure long-term growth for the Web, this consortium could establish AI standards that could provide an alternative to ensure Nvidia’s CUDA architecture does not become a monopolistic gatekeeper, allowing for other GPU vendors like AMD and Intel to compete on an even playing field.

• Develop and disseminate best practices for ethical AI use, ensuring that AI technologies are developed and deployed in a manner that respects human rights and promotes social welfare. This could mirror initiatives like the IEEE's Ethically Aligned Design, which provides comprehensive guidelines to prioritize human well-being in AI systems. The consortium could focus on establishing guidelines that ensure AI models, like those developed by companies like Meta, are genuinely open source, facilitating transparency and collaboration across the board.

• Organize and facilitate cross-industry and cross-border collaborations, helping to bridge the gap between technology and policy on a global scale. This would include forming advisory boards of members from diverse sectors—technology, law, ethics, and public policy—to ensure that legislation around AI is comprehensive and globally harmonized. Such collaboration could help avoid the fragmented regulatory approaches in the EU’s GDPR versus the US’s sector-specific privacy laws, promoting a unified global standard.

• Function as an industry liaison to help provide education so that governments can pass legislation that makes sense and not just one-offs that will frustrate innovations. For instance, the consortium could work closely with government bodies to educate lawmakers on the technical nuances of AI technologies. This would be akin to the International Telecommunication Union (ITU)'s role in standardizing and regulating global telecom networks and services, helping draft legislation supporting innovation while protecting consumer interests.

• Provide an industry trade organization for collaboration that doesn't provide the same risks that Apple and Microsoft are trying to avoid with OpenAI. This could involve creating a neutral platform for sharing AI developments and security practices, similar to how the Global Network Initiative (GNI) provides guidelines and advocacy for privacy and freedom of expression across the internet and telecommunications industries. This would help maintain an open and competitive ecosystem in AI development, ensuring no single entity, like Apple or Microsoft, can monopolize foundational AI technologies.

By adopting these mandates, the consortium would guide the development of AI technologies and foster a more ethical, transparent, and equitable global AI landscape.

In addition, I think an executive director with deep knowledge of AI, such as Dr. Fei-Fei Li, would be ideal to lead this organization. As the Co-Director of Stanford's Human-Centered AI Institute and former VP at Google Cloud, Dr. Li brings extensive expertise in AI research and industry applications.

Anyhow, these are the insights from someone who has run an industry consortium, the Node.Js Foundation, and been active in quite a few others, including the Linux Foundation and the Cloud Native Computing Foundation, as well as being a committer for the Apache Software Foundation. They’ve been foundational in bringing widely used foundational software into use sustainably.

Open Foundation Models

Along with the need for AI to flourish sustainably, I think we need a core of foundation models that are genuinely open source. This means the model weights should be accessible to all, enabling transparency, scrutiny, and customization. Notable examples include Mistral, Google, and IBM models, which exemplify true openness. In contrast, Meta's Llama models should be compelled to use a truly open source license or stop branding their models as open source. It’s my top gripe with what is otherwise a good initiative at Meta.

This core initiative is foundational (pun intended) and critical to diversifying the market and driving scientific discovery and overall technical advancement.

• Open Source Licensing - The model and its code must be released under a genuine open source license (e.g., Apache 2.0, MIT, GPL).

• Access to Model Weights - Model weights should be publicly accessible for viewing, modification, and distribution.

• Training Data Transparency - Training datasets must be openly available or documented with sources and preprocessing details.

• Comprehensive Documentation - Detailed documentation should cover the model’s architecture, training methods, performance metrics, and limitations.

• Community Support - Provide an active community and support channels to facilitate collaboration and address user issues.

• Ethical Use Guidelines - Include guidelines on ethical use and potential risks to ensure responsible deployment.

Open foundation models allow researchers and developers to build upon existing work, ensuring that innovations are shared and improved collectively. The open source community thrives on this principle, and extending it to AI models will enhance collaboration and innovation.

Collaborations for A Bright AI Future

We can ensure AI's sustainable and innovative future by focusing on these critical aspects—open foundation models, industry collaboration, and establishing a dedicated AI organization. These steps promote transparency and collaboration and safeguard against potential risks associated with AI development.

Feature Image Prompt

This week’s feature image was created in Midjourney.

/imagine A photography piece using a 35mm lens depicting an industry roundtable on the ethical implications of sustainable AI. The room is decorated with #CC3333 and #3399CC, with participants showing serious contemplation, under natural, diffused lighting. --ar 16:9

Prompt of the Week

ChatGPT to Runway ML Prompt

Video is one of the top ways to promote products and share ideas on social media. But most of us aren’t skilled at creating videos. Enter one of my favorite AI video platforms, Runway. It’s a high-quality, easy-to-use video generation tool, though I will tell you that it’s a little difficult to guide.

How To Use This Prompt

1. Copy this prompt into ChatGPT

2. Answer all the answers in the interview

3. Then, cut and paste the prompt into Runway.

# Objective

Create a prompt for Runway, a text-to-video generation model. 

# Instructions 
Do not echo the prompt. 

1. Conduct an interview based on the interview questions. 
2. Generate one question at a time, wait for the anser than proceed to the next question. 
3. When the interview is complete create a prompt for use with Runway based on the answers. 

# Interview Questions

What is the main setting of your video? (e.g., tropical rainforest, bustling city, serene beach)
Provide any specific details about the environment. (e.g., clear sky, sunset, dense fog)
Subject Focus:

Who or what is the main subject? (e.g., a person, an animal, an object)
Describe any specific characteristics or actions of the subject. (e.g., a woman standing, a dog running)
Camera Movement:

How should the camera move? (e.g., static shot, tracking, fly-through)
Describe any transitions or changes in perspective. (e.g., start with a close-up and pull back)
Lighting and Effects:

What kind of lighting should be used? (e.g., natural light, dramatic, silhouette)
Mention any additional visual effects. (e.g., lens flare, backlighting)
Style and Aesthetics:

What overall style should the video have? (e.g., cinematic, moody, vibrant)
Any specific visual aesthetics? (e.g., glitchcore, VHS home video)
Movement and Dynamics:

Describe the movement style of elements in the scene. (e.g., slow motion, dynamic motion)
Mention any specific types of movements. (e.g., emerges, ripples, unfolds)
Text and Titles:

Do you want any text or titles in the video? (e.g., title screen, captions)
Specify the style of the text. (e.g., bold, graffiti, neon)

Here’s my first attempt using this prompt. The quality of the video is pretty good, but it does not include all the specified subject matter (The generated prompt is in the description on YouTube.)

My two cents is that you may need to alter the prompts based on the first prompt to get better results. I am not sure there’s a benefit to using XML or Markdown, so just write using natural language. However, I would like to add spaces and line breaks in the second go-round to make the following easier.

I recommend 10Web for Wordpress users who want all the flexibility of the world’s most popular CMS, Wordpress, but with AI ease of use.

SAI Toolbox - Latest AI Tools and Services I am Evaluating

• AIEditor is a next-generation rich text editor specifically designed for AI applications. It offers advanced features tailored to streamline the creation, editing, and management of AI-generated content.

• LoomAI—I use Loom to create screencasts, but the new AI Suite, which is an add-on to Loom, has some very cool time-saving features, including Auto Titles, Auto Summaries, and others.

• Shortwave.com is a versatile platform designed to streamline email. It offers advanced email organization, real-time collaboration, and powerful search capabilities.

AI Toolbox Prompt

This week’s AI Toolbox image was created in Midjourney.

/imagine prompt: A comic book style panel inspired by Neal Adams, illustrating a woman with a serious demeanor, working on AI software at her computer. The scene is enriched with #CC3333 and #3399CC, creating a striking visual contrast, under intense, dramatic lighting. --chaos 20 --ar 16:9 --stylize 1000