Generation AI - What's Your Call Sign

?

AI Enthusiasts, Learners, Doers - we are excited to bring you this edition of The Atlas for a couple of reasons. First, because of YOU. Our goal is to build community; locally and globally - to help educate, to help others prosper by helping you stay at the forefront of all things AI.? Second, because our community is growing and we are proud to share that we are the fastest growing and largest community - where you might ask? - on planet Earth. Yes, you read this correctly. But it’s not about the stat, rather it’s about the potential of what we can collectively bring to each other.? So with that - read on, or should I say ‘rock on’!

In this installment of all things crafty - brought to you in partnership with Buzzy ! :

1. Open, Closed, Safety First - Fines are doubled in construction zones

2. Prompt This

3. Lead, Follow, or reimagine your career with AI

4. Scale, Fail and You’ve Got Mail

??? What’s Hot: - Open, Closed, Safety First - Fines are doubled in Construction Zones

My Pet Llama 2 just got loose - the artist formerly known as Facebook, just launched Llama 2 . While many are excited about the ‘openness ’ L2 can bring. The excitement overall - Llama 2 models are trained on 2 trillion tokens and have double the context length of Llama 1. Llama-2-chat models have additionally been trained on over 1 million new human annotations. Instruction Tune by Philipp Schmid at Hugging Face. Check it out.

Who’s missing out on the AI Boom - Editors at Linkedin just published an article off the back of Brookings Institute Report that many AI jobs are clustered in existing tech hubs in the US; San Francisco, San Jose, Los Angeles, etc. No need to fret - why? AI jobs right now are focused on building. As AI mainstreams, we assert that applications, workflows and knowledge of how to apply AI will 10x overtake the build hype and people will be engaged with embedded AI experiences with products and tools they love today, as well as some new ones.

Singapore signals, EU shapes the way, the White House advances - A month ago the EU declared the importance for governing AI (AIA AI Act ) and the implications to society. Just a few days back, the White House followed suit and aligned with 7 leading tech companies to help manage the risk and exposure of AI . Regulation and guardrails are fine, especially when web3/crypto woke many up to the perils of misguided growth - learn from the past, apply judgment for the future.

Responsible Use of AI - Insight Ventures recently published a good article on AI governance and responsible use - something we like to see. Why? When we all aim to understand how a common framework works, we abide by it, improve it and ultimately leverage it to do great work. Next time you get mad at road construction, remember, they are making it easier for you to - prosper.

?? Applying AI: - Prompt This

Long Live the Intuition Economy? - We love VentureBeat, though this notion that Intuition will outpace the knowledge economy , assumes what? We make sense of the world through eight ways of knowing: language, reason, sense, perception, memory, faith, intuition, imagination and emotion. In this case, knowledge will forever be our foundational set of learned experiences - written, read, seen, felt, smelt, tasted, heard. Without, Intuition is not well formed. In other words, when search arrived on the scene 20+ years ago, how many of us knew what questions to ask the funny search bar - many questions were derived from knowledge/experience, no? Fast forward to today - prompting - how many of you have that same feeling of - not sure what prompt to pose?? Long live experiential knowledge, our token based machines will thank us.

AI, The Final Frontier - Said nobody, but Sequoia did suggest that AI and the Frontier Paradox will continue to change. The (why now?) and (so what?) are compelling. (i) Technological breakthrough with LLMs becoming programmably extensible and with unbounded scaling. (ii) Human Language as the interface. We used to speak of the graphical user interfaces - ‘insert obligatory reference to re-watch Pirates of Silicon Valley’ and the profound impact it had with human+computer interaction (not the movie, but the GUI). We are on a xFactor trajectory with something most people know today - of which with AI, is translatable, relatable, and uniting - human language.

Enterprises of the 80’s called, they want their memo sticky note back - What’s amazing are the words we use to describe data - corpus, exhaust, the new oil, water, gold - add your favorite commodity. We used to joke, ‘with all this data you are storing, you are simply heating a data center’. The day to finally tap this stored energy and put it into use, is now. Folks at Glean, recently posted an article - Lessons Learned about building Enterprise AI Assistants - in short, RAG - Retrieval Augmented Generation allows for enhanced search + LLMs to ground in reasoning and synthesis. Simple, yet complex, it allows for organizations to tap their vast enterprise data to provide employees with actionable insights.

?

??AI Snippets: - Lead, follow or reimagine your career with AI

The University of Michigan published a university report on the implication, impacts and use of AI at the University. A well summarized report that, while higher ed focused, ANY enterprise can/should take note. Some quick nuggest of goodness.

(i) The GAIA Committee has developed an initial vision and recommendations for U-M to ensure responsible, ethical, legal, secure, equitable, accessible, and transparent development and use of GenAI. Recommendations include establishing a GenAI committee, leveraging GenAI for teaching and learning, assessing research use of GenAI, creating a research initiative, and providing secure and equitable access to GenAI platforms.

(ii) Establishing an AI Digital Commons

Office of the VPIT to work with GenAI committee to create a website providing strategy, policies, resources, and links for students, faculty, staff, and administrators. Includes science & capabilities of GenAI, guidance for instructors, and integrating AI tools.

(iii) Implications of GenAI

Skills and competencies for an AI-augmented world, best practices for writing pedagogy, teaching and learning methods, research practice, ethics and equity, and future impacts on university life and administration.

Executive Summary:?U-M community has mixed perspectives on potential, benefits and risks of Generative Artificial Intelligence (GenAI). Survey of 6,000 responses found that 60% of faculty and students and 40% of staff have used GenAI systems. U-M must responsibly leverage GenAI to further mission and support community while mitigating threats.

The Full Report can be found here

Who’s leading AI and Why. A friendly post, by a colleague of ours Bruno Aziza summarizing a view of which country is leading AI , published by folks over at Tortoise Media .? Interesting to see, though perhaps the real opportunity to suss out is where are AI communities being created - where are the hubs of expertise, interest, and collaboration happening?? We are excited to see as a global community of readers and doers of all things AI, the real magic aligns in a collaboration of countries unbounded. Share with us where you are seeing the collaboration firsthand and share it with us!?

AI and Culture: Every day, the news is all about the next LLM, performance and scaling. However, few are speaking about culture and the adoption of AI enabled technologies. Soon we hope to stop talking about AI, because it will just be ‘in’ what we use day-to-day (well, actually it’s in a large % of technologies we use today already). However, until then, we need to learn from those who've 'been there done that'. We are excited to give a pre-sneak to an upcoming webinar with the co-founder and CEO of Ivy.ai - Mark McNasby. As some can make elephants dance, others like the team at Ivy.ai are enabling generative AI through higher ed and doing so before generative AI became cool. It’s not strictly about the tech, but it’s about the demographic, engagement, use and the setting. Higher Ed being just one of main markets they serve, the learnings from their journey pre-pandemic till now, may help a few Enterprise's dance and quickly adopt AI - because guess what; the incoming workforce, powered by Ivy.ai are already pretty comfortable using these great technologies. Be on the lookout for an awesome session that you won’t want to miss!

Women Hit Hardest by AI Wave: Linkedin recently published a new article suggesting that Women may be hit the hardest in finding jobs as a result of the AI Wave . On the basis of employment and perhaps what jobs women have, the article suggests more women are in areas that have the potential for AI Automation to reduce jobs - customer support, service, etc.? While we know AI has potential for bias in model output, the belief that more women will lose their jobs due to AI is profoundly wrong. We think any job that has the potential for automation (gender aside) may be redefined, reduced or rather - reimagined with AI. The biggest opportunity we see is how routine tasks will be displaced, but those tasks will still require context, knowledge and guidance. It’s an opportunity for women or others to reimagine roles and opportunities as a result of AI. In short, catch and ride the ?? … you own it ??.

??The Stack: - Scale, Fail, You’ve got Mail

Hal 9000 is that you? -? Condor Galaxy - Cerebras and G42 The UAE Technology Holding Group announced the largest AI interconnected supercomputer . 4exaFLOPS and 54 million cores - it’s time to grab some popcorn. Topping out at 36exaFLOPS and providing a network of interconnected supercomputers aims to radically transform AI training at scale with a focus on healthcare, energy and climate - we are excited to see what Condor can bring.

Robots.txt - #Fail - in the ever existing space of ‘tagging’ Large Language Models (LLMs) bring a unique challenge to metadata management. A provocative piece by Filie Wiess over at Search Engine Land proposes why search engines like Google & Bing are aiming to eat their cake, while serving billions. Suggesting that crawling/indexing sites and harvesting content to train LLMs are one in the same - simply not. While banning the crawler from harvesting info for their LLM hurts my page rank, it feels anything but open. Enter Creative Commons and long live tagging, we’ll see how this plays out.

Challenges and Applications of Large Language Models - Jean Kaddour, Joshua Harris, Maximillian Mozes, Herbie Brandley, Roberta Raileanu and Robert McHardy, published an amazing research article on the challenges and applications for LLMs . In short, challenges include unfathomable datasets, tokenizer-reliance, high pre-training costs, fine-tuning overhead, high inference latency, limited context length, prompt brittleness, hallucinations, misaligned behavior, outdated knowledge, brittle evaluations, evaluations based on static, human-written ground truth, and indistinguishability between generated and human-written text.? These foundational understandings give great insight into various applications for LLM and considerations to take into account; ChatBots, Computational Biology, Programming, Creative and Knowledge Work, Law, Medicine, etc. While a takeaway summary would require its own paper, our encouragement is that this paper helps to spark an deeper understanding and collaboration across business, technical expertise to further refine pre and post conditions to scale LLM driven applications, solving some of the most complex, yet impactful opportunities many of us have dreamt could be approachable with technology. ? Grab a coffee (or two) and dig in.

The key to building a great product is a great team, not a great idea.

?? Midjourney prompt of the week:

"1970s, medium-closeup, Ektachrome, 25-year-old woman, curly red hair, freckles, vintage car, downtown Los Angeles, green silk jumpsuit, plunging neckline, oversized sunglasses, golden hour. --c 80 --v 5.2"

??When there are too many guardrails applied!

---

## Do you have an AI App? Showcase It Here . We are in the final pass of the marketplace, and are excited to share!

## Are you an AI content creator? Share Your Work with Us

## Sponsor with us like Buzzy did - reaching 775K Impressions and 275K video views in a prior post

Do you have news or events you would like to share? Contact Us. Reach our 500,000+ readers. Partner with us? [ info at genai.works ]