Thus far, if the truth be told, I’ve had mixed experiences with large language models (LLMs) and generative artificial intelligence (GenAI) as personified by ChatGPT. I started, as many people do, by asking ChatGPT questions about things I knew nothing about, receiving responses that appeared to be reasonably impressive.
The reason for my emphasizing the word “appeared” in the previous paragraph is that I next asked questions about things I did know something about, only to discover that many of ChatGPTs responses were missing key elements or, in some cases, were… well… let’s say “perhaps not as accurate as one might have hoped.”
As an aside, I must admit that I’ve had significant success with GenAI image generators like Stable Diffusion, and I’m hoping to have a lot more success in the not-so-distant future.
Do you remember my column Generative AI Is Coming to the Edge? In that column I noted that I’m in the process of writing an autobiographical book describing my formative years from 0 to 11. Titled The Life (of a Boy Called) Clive (It Rhymes with Five), I’m planning for the tales in this tome to be accompanied by hand-drawn black-and-white pencil sketches. I’m envisioning these as being like the illustrations in Winnie the Pooh, except mine will feature me and my teddy bear buddy rather than a teddy bear and his piglet pal.
The only problem is that I’m no good at creating this sort of thing and I cannot afford to pay an artist to do it for me. The solution may be GenAI. You can achieve extraordinary results by iteratively refining your requests, but this can be a bit of a puzzle, and it does require time and perseverance. Fortunately, my wife (Gina the Gorgeous) is awesome when it comes to perseverance and puzzles. Gina had completed The Room on her iPad while I was still trying to wrap my brain around the early enigmas.
The reason I mention all this is that, when I penned the aforementioned column, the guys and gals at Kinara (which is derived from the Hindi word for “Edge”), were working on a USB stick-based edge GenAI image generating solution featuring their Ara-2 AI Processor. This is something Gina could plug into her notepad PC and use to happily beaver away generating the images for my book. The chaps and chapesses at Kinara made mention of the fact that they might send me one of these bodacious beauties (no promises). If so, I will report back in detail in a future column but, as usual, we digress…
Having said all the above, I recently did have success with ChatGPT. I was supposed to generate a proposal for a project. I’m not very good at the business side of things. I fear I prevaricated, and I left things far too late. I ended up with only about an hour remaining before I was supposed to deliver the proposal. I wasn’t sure where to start. I called my chum Joe Farr in the UK. He suggested using ChatGPT. He told me to simply tell ChatGPT what I wanted to do, employing a chatty style as though I was writing an email to a friend. I did so and… blow me down… it worked!
I didn’t use ChatGPT’s response directly—I cut it down and edited it substantially—but the result was to generate something in a few minutes that would have taken hours had I been obliged to craft everything by hand. Suffice it to say that this certainly opened my eyes to GenAI possibilities.
Why am I waffling on about this here? Well, I just saw something very similar with respect to generating the schematic for an electronic product.
I was chatting with Tobias Pohl, who is Co-Founder and CEO, and Rob Telson, who is VP of Global Sales, at CELUS. Tobias told me that this all started when he and his co-founders were working on a project design. They realized that they were finding it extremely difficult to decide on the right components. They likened this process to looking for the proverbial needle in a haystack.
Finding the right component can be like looking for a needle in a haystack (Source: CELUS)
The way they describe this is that millions of design engineers invest countless hours wading through data sheets trying to find the right components for the job. Meanwhile, component manufacturers and distributors/suppliers end up investing vast resources in the form of field application engineers (FAEs) and technical support trying to persuade people to pick their parts. This led to the founding of CELUS and the creation of the CELUS Design Platform, which acts as a bridge between these two domains.
Building a bridge between designers and component suppliers (Source: CELUS)
CELUS was founded in 2018. They are headquartered in Munich, Germany; they have a design center in Porto, Portugal; and they are launching their commercial headquarters in Austin, Texas. They currently have 90 employees; they expect this number to rise to a little over 100 by the end of 2024; and they are anticipating having between 130 and 150 employees by the end of 2025 (on the off-chance you are looking for a job yourself).
CELUS Design Platform is 100% cloud-based, which means you can use it on anything from a tablet computer to a laptop to a workstation. The way to think about the CELUS Design Platform is like a trusted advisor. The easiest way to wrap your brain around this is to bounce over to their website, register, and start playing with it. I just did so myself. I signed in with Google, which took no more than a couple of clicks.
There are multiple ways to use the platform. There’s also a handy-dandy self-guided tour, which will have you up-and-running before you know it. You can start by selecting one of the existing reference designs, which you then iteratively tweak until it meets your requirements, or you can commence a new project from scratch.
Assuming you select the “New Project” option, you begin by stating in natural language just what it is that you are hoping to achieve, like “I want to create a microcontroller-based thermostat that I can monitor and control using Bluetooth,” for example. You can also specify things like regulatory requirements, environmental conditions, preferred manufacturers (if any) and manufacturers you wouldn’t touch with a barge pole (if any).
A simple drag-and-drop interface allows you to drag what they call “design functionalities” from a graphical library on the left and drop them on your architectural canvas in the middle. For example, you could drag a temperature sensor and a microcontroller and then connect them together. If you have any specific requirements, like the range and resolution of the sensor, for example, you can right-mouse click on that block to access a features window. If you select a specific interface, such as I2C, for example, then this will form part of the component selection criteria.
When you are ready to rock-and-roll, you click the “Resolve” button, and the platform uses its smart algorithms and GenAI to search through myriad component combinations (including perusing and pondering their data sheets). The outputs include the architecture for the circuit, using only components that are actually available on the market, along with ECAD native schematics for the schematic capture tool of your choice (you’ve already told the platform which schematic tool you favor).
Turning ideas into component selections and schematics (Source: CELUS)
Once again, the way to think about the CELUS Design Platform is as a trusted advisor. The design functionalities are known as CUBOs. We can think of these as knowledge-enriched, all-inclusive, digital datasheets that contain all the vital information about an electronic component needed to choose it and use it in a design. Your AI assistant is called CUBOt. You can ask CUBOt natural language questions like, “Why did you choose this component” or “why did you opt for this interface type” or “what happens if we…”
Rob told me that, in his previous position, he was part of a team developing an edge AI box. He said the team met every Monday morning to talk about the design and component selection and integration and interfaces and power consumption… and every week that they met something had changed. One week it was the microcontroller, the next it was the interfaces, the next… Rob said that it took them six months to get to a completed design, only to find it included some parts that weren’t actually available. Rob says that the work that took them 6 months would have taken only a few hours if they had had access to the CELUS Design Platform. It’s no wonder that Rob came onboard at CELUS as soon as he met Tobias and became aware of the CELUS Design Platform.
Now, I tend to think about things from the view of a hardware design engineer, because that’s what I am deep down. However, it turns out that the CELUS Design Platform is also of interest to component manufacturers and component distributors/suppliers. In fact, there are three engagement models that are applicable to these entities.
Manufacture and component distributor/supplier co-branding (Source: CELUS)
The entry-level model involves taking parts from a manufacturer or component supplier and adding them into the platform. The next level up is co-branding, whereby the manufacturer, distributor, or supplier takes the CELUS Design Platform and makes it available on their own website as a design tool. Finally, some companies are approaching CELUS saying that they want a fully customized “White Label” solution.
The idea is that these solutions let external users and internal personnel (like FAEs) use the platform to select appropriate components and access information about those components, including how they will work (or not) with other components. In the case of the co-branded options, the results will, of course, be biased toward each companies’ offerings and product lines. In some cases, there will be both external and internal co-branded platforms, where the internal platform allows FAEs and internal users to gain access to components and data that are not currently available to external customers.
Phew! There’s a lot to consider here. All I can say is that in the same way using ChatGPT makes you think about things in a different way, so will the CELUS Design Platform open your eyes to a new world of possibilities. Also, that it’s well worth your time to visit the CELUS website and take this little beauty for a spin. If you do, I’d love to see your thoughts in the comments below.
One thought on “From Thought to Circuit in Record Time with AI”