Microsoft this week announced the public preview of Python in Excel, which will allow advanced spreadsheet users to combine scripts in the popular Python language and their usual Excel formulas in the same workbook. The feature will first roll out to Microsoft 365 Insiders as part of the Excel for Windows beta channel. While Anaconda, Microsoft, and consumers are clear winners, there are more than a few technology companies that could lose out. Of course, this will largely be determined by Microsoft’s bundling and revenue strategy.
GPT-4 large language model supports multimodal (images, text, sound, and video) that are the next step closer to general AI with some improved security controls. At a rumored 100T parameters, GPT-4 would be over 500 times larger than GPT-3. With America’s AI market size estimated at $119.78 billion. Almost every business and over 250 U.S. Startups are developing Generative AI applications, pursuing Large Language Models, Machine Learning, Neural Networks and Digital Twin technology.
Last week, more than 1,000 scientists and tech leaders, including Elon Musk, signed an open letter calling for a pause in the race to develop more powerful artificial intelligence models. The letter channeled a certain dread that many feel about this fast-changing technology. It also became a lightning rod for criticism from both AI advocates and skeptics.
The primary fears are hallucinations, bias, disinformation, and productivity improvements that result in job loss, loss of agency, and loss of control. The idea that we can regulate or control technology seems unrealistic on a national level and unimaginable globally, so we may have to rely on Risk Management Frameworks, auditing, and the old On/Off switch until we can train AI to police itself from all of the maladies and disorders that it has learned from our societies.
Simple truths 1. All technology can be used for good or evil. 2. Any significant change will result in unintended outcomes. 3. Those postulating fears from technology are typically heavily invested and have the most to lose. 4. AI won’t entirely replace humans, but humans with AI will replace humans without AI. 5. Whether AI nets out on the side of humanity or not, it will accelerate everything.
Last year Google Research announced a vision for Pathways, a single model that could generalize across domains and tasks while being highly efficient. An important milestone toward realizing this vision was to develop the new Pathways system to orchestrate distributed computation for accelerators. In “PaLM: Scaling Language Modeling with Pathways”, Google introduce the Pathways Language Model (PaLM), a 540-billion parameter, dense decoder-only Transformer model trained with the Pathways system, which enabled us to efficiently train a single model across multiple TPU v4 Pods. They evaluated PaLM on hundreds of language understanding and generation tasks, and found that it achieves state-of-the-art few-shot performance across most tasks, by significant margins in many cases.
The journal-of-applied-research-in-memory-and-cognition is a valuable resource for software engineers interested in moving beyond pattern recognition into cognitive intelligence. Some cognitive architecture frameworks are Soar, CLARION, and EPIC and ACT-R. The real value will be in how we interact with AGI.
