Methinks Elon should focus on fixing Autopilot in Tesla than worrying about and suing OpenAI for achieving AGI!

References:

I was in the Chicago burbs, and it was bright and sunny in the morning. Driving west, my Tesla failed to recognize several red lights due to glare! Some other points:

• Auto park is an issue now that ultrasound sensors and radar are disabled.
• Snow and rain blocks camera view thus disabling the vision-only auto features.
• The wipers that auto clean are now driven by cameras (don’t have the pressure detection from Auto Pilot 0 and like in other cars), and sometimes move too fast or too slow.

I know Musk’s contention is that humans have two eyes to facilitate driving and 8 vision cameras should be sufficient. But the intention should be go beyond just replicating human functions to meet with an objective function.

Also, AGI is a misnomer, I think it should be called Artificial Collective Intelligence, wherein a distribution of human and machine “nodes” coordinate to enact intelligent solutions for our daily endeavors.

In a recent webinar, the discussion was on the power and possibilities of 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐫𝐭𝐢𝐟𝐢𝐜𝐢𝐚𝐥 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 and 𝐄𝐱𝐭𝐞𝐧𝐝𝐞𝐝 𝐑𝐞𝐚𝐥𝐢𝐭𝐲 technologies in industrial environments. Key takeaways from the discussions were:

• Rapid technological advancement: Technology evolves faster than cultural adaptation, necessitating proactive measures to integrate these innovations responsibly.
• Responsible AI adoption: We must act as both guardians and developers of AI technologies, ensuring their ethical and beneficial use.
• Iterative approach to adoption of AI: A “nail it, then scale it” mentality is crucial, emphasizing experimentation and refinement before widespread implementation.

We, at Numorpho Cybernetic Systems (NUMO) are building a process engineering platform based on LLM, Simulations and Computations and an inference engine based on GenAI. Our goal is to effect upstream, midstream and downstream activities using AI and XR in conjunction with Digital Twins and Digital Threads to concert end-to-end processes. We call it Actionable Intelligence where we intelligently convert meaning to motion to enable processes not only in the shop floor but also in product development and after-market.

We recognize the transformative value that AI brings to the table. In our case, we intend to utilize its power to enable process engineering, connecting the dots between people, processes and platforms to simulate, emulate, interact and enact actionable intelligence by converting meaning to motion. We intend to achieve this securely, ethically, responsibly and instituting explainable mechanisms utilizing science, math and engineering to train and validate our responses every step of the way.

Thus, Narrow Intelligence specific to a bounded box domain is what we plan to achieve first rather than complain about General or Super Intelligence that others are purporting to have.

Multimodality is also key for correct response – enabling stitching together different sensory inputs and their perception models to effect a solution is key to getting thing right. This is where LLMs like Google’s Gemini is scratching the surface of next generation intelligence, and we plan to utilize it for our underpinnings.

LLMs Review

Here is a good article comparing different LLMs:

Which Generative AI Is the Best? ChatGPT vs. Gemini vs. Pi vs. Claude 2 | LinkedIn

We, at Numorpho Cybernetic Systems (NUMO) have been utilizing these LLMs separately, in conjunction or with the same prompts to make sense of the variations in the output, and also to build our own knowledge base for process engineering.

Here is our summary of the four:

• OpenAI #chatgpt was our original go to model but we have found it to be less informative lately. Its inclusion in Microsoft‘s #bing and #copilot is good but there is a lot of extraneous information that needs to be weeded out.
• We like Google #bard , now #gemini because of the 3 optional responses it gives for each prompt. But recently the imaging feature has been disabled due to it being too liberal. I don’t think this is an issue for us in what we do but have noticed it when we did literature review of some controversial books like John Steinbeck’s Of Mice and Men, and George Bernard Shaw’s Caesar and Cleopatra.
• Anthropic #claude is amazing and we are looking at its version 3 to see how much better it is.
• Inflection AI#PI is very conversational and it is like talking to a effervescent human. We used it a lot to understand the ramifications of the “Coming Wave” of AI, synthetic biology, nanotech, quantum and other emerging technologies.

Our goal is to enable themed prompt engineering to facilitate product innovation, process automation, coordinated integration between systems internal and external to the enterprise, and what we call actionable intelligence to convert meaning into motion.

Most of our customers, us included, would like a closed box solution for prompt engineering which will be the lead to doing complex simulations and engineering future products and solutions. Our endeavor is to create LLM in a box with the appropriate knowledge base for that domain with past curated information that would provide the cybernetic basis for communication, control and using feedback loops to prove and improve our platform for process engineering.

Thusly, we are intently following companies like NVIDIA with their #Omniverse, #DGX and other solutions like Neuromorphic Processing Units (#npu).

For this we are also looking at #mistral, #llama, #orca and others to create a #MemGPT based on multi-modal LLMs synthetically training and learning from each other so that issues of hallucinations /confabulations – #halluconfabuphobia, context length and others are mitigated.

Here is sound advice from Jeff Winter:

From CPUs, to GPUs to NPUs

𝐍𝐕𝐈𝐃𝐈𝐀’𝐬 𝐣𝐨𝐮𝐫𝐧𝐞𝐲 𝐟𝐫𝐨𝐦 𝐚 𝐬𝐭𝐚𝐫𝐭𝐮𝐩 𝐟𝐨𝐮𝐧𝐝𝐞𝐝 𝐢𝐧 𝐚 𝐃𝐞𝐧𝐧𝐲’𝐬 𝐫𝐞𝐬𝐭𝐚𝐮𝐫𝐚𝐧𝐭 𝐭𝐨 𝐚 𝐠𝐥𝐨𝐛𝐚𝐥 𝐭𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲 𝐩𝐨𝐰𝐞𝐫𝐡𝐨𝐮𝐬𝐞 𝐬𝐡𝐨𝐰𝐜𝐚𝐬𝐞𝐬 𝐭𝐡𝐞 𝐭𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐯𝐞 𝐢𝐦𝐩𝐚𝐜𝐭 𝐨𝐟 𝐆𝐏𝐔𝐬 𝐨𝐧 𝐦𝐨𝐝𝐞𝐫𝐧 𝐜𝐨𝐦𝐩𝐮𝐭𝐢𝐧𝐠. 𝐉𝐞𝐧𝐬𝐞𝐧 𝐇𝐮𝐚𝐧𝐠’𝐬 𝐥𝐞𝐚𝐝𝐞𝐫𝐬𝐡𝐢𝐩 𝐡𝐚𝐬 𝐛𝐞𝐞𝐧 𝐢𝐧𝐬𝐭𝐫𝐮𝐦𝐞𝐧𝐭𝐚𝐥 𝐢𝐧 𝐬𝐡𝐚𝐩𝐢𝐧𝐠 𝐍𝐕𝐈𝐃𝐈𝐀 𝐢𝐧𝐭𝐨 𝐚 𝐤𝐞𝐲 𝐩𝐥𝐚𝐲𝐞𝐫 𝐢𝐧 𝐀𝐈 𝐚𝐧𝐝 𝐆𝐏𝐔 𝐭𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲.

https://www.linkedin.com/posts/joeoquist_generativeai-genai-ai-activity-7167518603519463424-C5Ur?utm_source=share&utm_medium=member_desktop

Wonder what the next wave of NPUs (Neural/Neuromorphic Processing Units) is going to bring….

NPUs are a new breed of processing units, inspired by the intricate architecture of the human brain. Unlike GPUs, which excel at parallel processing of large matrices, they strive to mimic the interconnected web of neurons and synapses, allowing for more natural and efficient handling of tasks that involve learning, pattern recognition, and adaptation.

Neuromorphic computing aims to emulate the structure and function of the human brain, using specialized hardware called NPUs. Unlike GPUs, which excel at parallel processing of large data sets, NPUs are designed for low-power, high-efficiency computation with a focus on mimicking brain-like neural networks.

Our company name, Numorpho Cybernetic Systems (NUMO) has its basis from Neuromorphic Engineering – distributed computing using large, connected databases and utilizing cybernetic control and communications and the ability to use feedback to morph solutions to enable actionable intelligence to convert meaning to motion.

We posit that future AI (and Quantum) computing will need a different architecture than current transistor chips and will be researching on the utilization of NPUs for our basis for what we call Existential Intelligence that will progress AI beyond its current brittle manifestations.

A multi-modal hybrid processing approach would be needed for future solutions that would appropriate the use of CPUs, GPUs and future NPUs to enable our next generation products and services.

NITIN UCHIL Founder, CEO & Technical Evangelist
nitin.uchil@numorpho.com

REFERENCES

• 20240101 – Happy New Year – EVERYTHING CONNECTED – Numorpho’s Book of Business

BELOW THE FOLD

Suggestions by Gemini:

• Numorpho Cybernetic Systems (NUMO): You describe building a process engineering platform leveraging Large Language Models (LLMs), simulations, computations, and GenAI-powered inference engines.
• Goal: This platform aims to optimize upstream, midstream, and downstream activities by integrating AI, XR, Digital Twins, and Digital Threads across the entire process chain.
• Actionable Intelligence: This concept translates meaningful insights into actionable steps, driving improvements not only on the shop floor but also in product development and after-market support.

Potential areas of focus

• Specific use cases: Highlighting concrete examples of how NUMO’s platform addresses challenges in different manufacturing segments would strengthen your message.
• Impact on stakeholders: Demonstrating how your platform benefits various stakeholders (e.g., engineers, operators, managers) can broaden its appeal.
• Ethical considerations: Briefly addressing how NUMO ensures responsible AI development and deployment would align with your emphasis on being “protectors” of AI technologies.

By incorporating these suggestions, you can effectively communicate the transformative potential of your platform and position NUMO as a leader in AI-driven advanced manufacturing.