Adel Chaveleh

This is how Large Language Models (like ChatGPT) work: GPT stands for “Generative Pre-Trained Transformer”. Let’s break down what this means. Generative: this means the model can write (or “generate”) new text. As opposed to simply analyzing or processing inputs, it can output totally new, unique text. Pre-Trained: when you hear about AI being trained on massive data sets, this is what they are talking about. In the pre-training phase, they give the model tons of data, which in the case of LLMs, it uses to learn the patterns of language as well as general knowledge. Transformer: this is the key that has made all modern LLMs and the recent explosion in AI possible. In 2017, a team at Google introduced transformers in a landmark paper called “Attention is all you Need.” Transformers allow the model to analyze every word in a sentence simultaneously in order to understand which ones are most important. (Previously language models tried to analyze words in sequence and wound up having memory issues). — Ok, so how does this all work? At the beginning they start with a shit-ton of data. They can’t just dump it all on the Large Language Model, it needs to be cleaned up and formatted properly. This takes a lot of time. Next, all the words need to be turned into numbers. Computers don’t actually understand words, so the words are broken up into “tokens” and assigned a unique number identifier, which the computer can use to reference that word. They then use these numbers to create an insane multi-dimensional map of the entire language. And when I say multi-dimensional, I’m not talking about 3D space like our puny human brains can comprehend, I’m talking about THOUSANDS of dimensions. They place these word-numbers (“vectors”) into a multi-thousand dimensional space map. This is essentially a meaning map that allows the computer to comprehend language. All existing words are placed somewhere on the map based on their meanings and their relations to each other. This is what it uses to predict words. The next phase of training is basically playing a prediction game where the model is given examples of sentences and asked to guess what the next word will be. If it guesses wrong, its “weights” are tweaked and then it tries again. This happens over and over again billions of times, until it gets really good at predicting words. Once it reaches a good place, its weights are fixed. There’s an additional layer of “fine-tuning” for personality, telling it what its role is, what style it should communicate in, and setting some parameters around how it should “behave”, but this is the basis of the mechanism it uses to generate text in realtime when you send it a prompt. — I’m curious if this makes sense or is still too technical? My goal is to make technical stuff like this more accessible, so would love any feedback if you’re feeling like: “what the hell are you talking about?”

3

Fortune Favors the Bold! Jimmy Treybig was the Founder of Tandem Computers, a great Texan, an alum of Rice University, and a passionate mentor. He started the company when the industry said there would never be another computer company after IBM.  Great Jon Y, Asianometry video on the history of Tandem Computers, from their innovative, fault-tolerant designs to their impact on ATMs and beyond. Tandem Computers redefined reliability amid evolving technology. Customers included US Navy, FedEx, AOL, Dell Technologies, Citi, AT&T, Nasdaq, the Tokyo Stock Exchange, Wells Fargo, John Deere, Safeway, Target, . . . https://lnkd.in/gvfi8uEw Something Ventured: The story of how a few individuals built an industry that would go on to finance nearly all of the major advancements in technology in the past 50 years, with companies such as Atari, Genentech, Apple, Intel Corporation, and more. https://lnkd.in/g-t3_RHr The Bold will Build Our Next Golden Age. Autonomy Institute, ATRIUS Industries, Inc.

26

3 Comments

Disney paid $1 billion to OpenAI to stop fighting an AI war they can't win. The deal has been hailed as a breakthrough for AI storytelling. I see it differently. This is just the inevitable monetization of AI's impact on IP - and luckily it seems to be happening sooner than in other IP piracy cycles. I've watched this pattern play out for decades. New tech emerges. IP gets used without permission. First it is ignored because it is too small. Then IP owners try to stop it, usually with lawsuits. They mostly fail. Then after it is long clear they lost, they cut a deal. Fighting rarely works. The technology and consumer adoption happens whether you participate or not. Disney isn't looking for a 20x return here. A billion dollars is roughly 0.2% of OpenAI's valuation. This is about securing a seat at the table. My guess? It's a round trip deal. Disney invests, OpenAI provides favorable licensing terms. Smart positioning, not a moonshot. IP owners need to accept a harder unfortunate truth: AI will use your intellectual property whether you want it to or not. I think of it this way: Disney couldn't stop us from drawing Mickey Mouse because we have pencils. (Mickey literally might be the only thing I can draw decently.) But the difference now is scale. AI enables instantaneous mass creation and distribution of derivative content. Legal action won't change that underlying capability - just like piracy it'll just keep popping up whack-a-mole style. I suspect ultimately the future looks a lot like what YouTube and Facebook already built for music and video IP on their platforms. Copyright protection systems that identify IP usage and let the owner decide: take it down, or take the monetization. The music labels learned this lesson. They do far better collecting revenue behind the scenes than fighting every infringement in court. AI will follow the same path. More IP will get licensed to these models until eventually the technology has built-in systems to identify pre-existing IP and route monetization to whoever owns it. This isn't the end of IP protection. It's the beginning of a new era for it. Companies that figure out how to monetize AI usage of their content will come out ahead. Those still fighting the last war will see their IP used anyway, with nothing to show for it. Just look at Viacom and their decade long lawsuit with Youtube. The question isn't whether AI will use your IP. It's whether you'll get paid when it does. (Witness the 10 seconds it took me to make the below image.)

66

44 Comments

Discover why Central Texas companies are turning to direct hire staffing in Austin to build stronger, more reliable teams. This video explores the long-term benefits of direct hire, including better retention, cost savings, and access to top talent.   Learn how GSG Talent Solutions helps Austin businesses hire smarter and grow faster.   #DirectHireAustin #AustinStaffing #GSGTalentSolutions #StaffingAgencyAustin #TempAgencyAustin

5

Exciting to see midstream operators like Oklahoma Gas Gathering leading the way. Partnerships like this show how Bridger Photonics data and technology can help raise standards across the industry. Check out our press release here: https://lnkd.in/gXGhE5Ei

25

Unlocking Stranded Assets: Transform Your Closed "Sour" Wells into Producing Assets 🛢️ Viro Petroleum and Energy has developed an innovative H₂S removal technology (U.S. Patent 11,813,551) designed to reactivate closed "sour" oil wells safely and profitably. https://lnkd.in/gwyx2e3D

2