Corporation Blog by AICU President Akihiko Shirai on November 14, 2023 Japan time: a look back at OpenAI DevDay and AICU's future plans for activities in the generative AI market, the expected market engagement for generative AI, and AICU's vision and core competencies for the AICU's vision and core competency, "Creating People Who Create," will be presented.
On November 6, a number of new features were released at OpenAI DevDay.
We have translated and analyzed them in Japanese as soon as possible, so please take a moment to read them.
https://corp.aicu.ai/ja/openai-devday-20231106
The biggest impact of this release is the release of GPT4 Turbo and its lower price.
While the lower price per API is great for communication AI app developers like us, we don't want OpenAI's business to become unstable. I imagine it will take some time to complete the release of ChatGPT5, but the premature release, the premature low price, and the fact that the chatbot “is not stable” are concerns (as we can tell from running chatbots).
Another notable new feature is improved function calls, "JSON mode" and "AssistantsAPI”.
The JSON response implemented with function calls was a shocking feature for chatbot development, with literally unlimited possibilities, but there was a blurring of how the actual function was called, for example, the function called for a question with multiple answers. The same is true for the JSON mode; if it could be designed to always return JSON, it would make the connection between systems more reliable. Then there is the "AssistantsAPI". This has the potential to make a very fancy demo, and OpenAI itself is hoping to get it buzzing on social networking sites.
I expect that we will see some interesting things for a while, but I have a feeling that from an actual developer's point of view, "we won't see that much diversity". This is because APIs are programming interfaces for application development, and if they are provided as APIs, the UX (user experience) created by their functions, price, and response speed will be generalized to some extent. On the other hand, examples such as "Goblin.tools," which I introduced last week, are excellent. Not only are the functions, response speed, and UX extremely sophisticated, but the fact that it delves into the prompt art of "breaking down ToDo" that no one seems to notice is commendable. What's more, developer Bram De Buyser (whom we interviewed) has made it possible to offer this web version for free through smartphone app sales.
It is important to repeat the obvious loop of not only the prompt art, but actually making it, making it practical as a subject, and then improving it in a more cost-effective way.
The same goes for the company AICU. Currently, we have just started and the only activity that comes to the surface is the media business, but we are getting very interesting client engagements in development, consulting, and book production, especially for the "AIDX Workshop".
We are getting a lot of serious inquiries, especially from companies that can compete globally at the top of their industries.
(This story is a blend of several companies' current situations.)
Many of these companies do have a good development team. We are starting to see a glimmer of "AI users" in our company as well, including OpenAI. There are likely to be many solutions and proposals from larger companies. However, even if OpenAI-based services progress rapidly and we can actually develop something as a B2B enterprise IT project with a big budget, it will be difficult if OpenAI shakes our price base and development base. It is difficult to keep up with the speed of releases every three to six months, and OpenAI itself destroys the modules that you have worked so hard to create from the design.
After all, there will probably be a period of time when nothing can be built in a large sitting group.
While you are turning around designs and PoCs, the infrastructure is being updated.
While we are turning around research, design, budget acquisition and PoC development, it is becoming clear that we need to build a small team to create a technology PoC with a high skill average and decision speed. If you don't build a small team, the revenue from the resulting service will always exceed the cost of development in a large sitting group.
The same is true in the non-development BizDev field. The base cost, just like ChatGPT's monthly fee or Adobe's monthly fee, and the cost of researching new services that keep coming up, can an individual afford it? The key is whether or not you can afford the base cost, like ChatGPT's monthly fee or Adobe's monthly fee, and whether or not the "added value created by using the service" is more than the monthly fee.
In particular, I feel that projects that sell "labor saving with AI" are destined to become increasingly thin. Unit costs are getting lower and lower, and people don't pay steady money for surprises. It will be harder and harder to get a higher selling price from someone who says, "I don't know, but it's great," or "It'll be cheaper if we use AI to save labor.
I, of all people, have been using Adobe products for 20 years since Photoshop 2.0, and these days, with only Adobe products (Illustrator, Photoshop, Premiere, Acrobat), knowledge of creative AI, and my skills, the cost of outsourcing is almost zero. I would rather spend that budget on excellent work. In fact, if I had that budget, I would certainly be able to add value by teaching talented student interns how to use the products.
Therefore, our vision is to "create people who create”.
Let me explain in another direction. This is the perspective of "creating an attractive workplace through AI development. Recently, many companies have been seeing employees who are proficient in using ChatGPT, etc. (AICU is not only using ChatGPT, but also image generation AI.
(Although AICU is not only using ChatGPT, but also image generation AI and blogging to the society like every day)
(1) A highly skilled "AI user" will emerge.
(2) That person will make more and more improvements / or make PoC proposals.
So far, so good. I sympathize. However, it seems that recently the users of generated AI have been separated from the rest of the company.
The speed of evolution is just too fast. The speed of knowledge cannot keep up, so internal organization cannot keep up either. Only those who are "a few people with a lot of skills" are "doing what they do because they are good at it," and no matter how amazing the PoC they create, it does not resonate or move forward as a business. On the contrary, we are often asked by people in the legal and PR fields, "Isn't generative AI illegal? or "You can't use this for commercial purposes, can you? Yes, innovation has few people who understand it.
(3-1) Few people in the company understand → (1) will leave the company
(3-2) If activities to increase the number of people who understand within the company can be done, the average will increase, but what (1) wants to do will not progress very much.
(3-2) If we can increase the number of people who understand, the average will increase, but what (1) wants to do will not progress very much.
Here
(3-3) AICU is able to raise the average by providing advocates and workshops.
AICU can contribute to raising the average level of your company’s AI use, comprehension and skills, not only technical but also legal/social communications.
In the era of generative AI, the terms ’advocate’ and 'advocacy’, originally meaning a spokesperson or defender, have taken on a multidimensional significance. If we were to describe this in terms of the Net Promoter Score (NPS), it could be said that the progression is from positive > supporter > advocate. Think of it as going beyond just loving your job; it's about being an ardent supporter.
In our internal workshops, employees are always expected to act as organizational members. Whether it's about innovation, cost, mission breakdown, or challenges, they face a constant push and pull in the company, a phenomenon they are already familiar with. However, everyone knows that this isn't the essence of the matter.
Our paid VIP 'AIDX Workshops' offer management a brief, immersive experience with LLM and image-generating AI, and service development. By turning zero into one, we greatly enhance understanding and significantly reduce meeting times. These workshops are not one-offs; we conduct them for all employees, too.
But remember, we are not just 'AI product vendors.' Our service is "creating people who create," not just using tools.
We advocate for hard-working employees and move them from a state of hesitation to being active creators, clearing ethical and legal ambiguities. With the next Windows Update bringing Bing search and DALL-E, it's essential to update our understanding of ethics and law. We aim to enhance common knowledge among employees to reduce legal costs. Using ChatGPT can expedite interpretations and decisions in as little as five minutes.
Then we do external advocacy to global reach. AICU Inc consistently communicates through open-source platforms and commercial blogs, articulating 'the societal average' in the fast-moving stream of generative AI. When drowning in information is no longer a concern, decisions become faster and more cost-effective.
Our 'AIDX Workshops' may seem roundabout, but by purchasing these experiences, employees reach a baseline understanding and reduce the likelihood of talent leaving the company. Rather than falling for dubious AI products and seminars, understanding open-source activities helps in organizing internal study groups and meeting exceptional AI developers and partners.
Founded in 1994, this pioneering digital creation institution has a history of about 30 years, producing numerous creators worldwide. Our CEO, Dr. Shirai, manages the 'Creative AI Lab (CAIL)' in the university's graduate program. https://dhgs.shirai.as/
This lab's manifesto is to carry out the PDCA cycle of "creating services and disseminating them through technical publications" in the age of Creative AI. The lab aims to bring together people who have been active in their respective industries (IT, banking, artists, etc.) and guide them through the steady and muddy process of creating value by "creating people who create," "developing services for as little as 500 yen a month," and "selling their creations to others, even if they are technical fanzines. By providing hands-on guidance in the value-added work of creating value, we are not only raising DCM to a "professional level" rather than a school level, but we are also developing human resources who will lead the industry for years to come. AICU's AIDX Lab has deployed this from methodology to actual business.
If you have students in your area, please let us know. Overseas students are also welcome.