Navigating the A.I. Era: How Riversoft from Taiwan Harnesses the Power of Generative A.I.

Exclusive Interview with Claude Shen, Chief Technology Officer of Riversoft Inc.

In a world where Generative A.I. has become ubiquitous, Riversoft‘s CTO, Claude Shen, discusses how the technology can be effectively incorporated into the business world. With A.I. market projected to reach nearly 2 trillion U.S. dollars by 2030, Claude explores the potential of ChatGPT-like models and emphasizes the importance of understanding their limitations. Discover how Riversoft aims to delve deeper into the power of A.I. to provide customised solutions while preparing society for a future where A.I. plays an increasingly prominent role.

Game Changers with Promising Futures

The emergence of Generative A.I. and applications like ChatGPT has transformed what was once a seemingly distant technology into an integral part of our daily lives. People from all walks of life, including researchers and laymen, were both shocked and impressed by its capabilities. The question now arises: How can we harness its potential and effectively incorporate it into the business world and our daily work?

“It’s a question of a ubiquitous technology that penetrates every industry. So with this generic model that can easily be adapted into all different kinds of industry”, according to Riversoft’s Chief Technology Officer, Claude Shen.

Claude is a Natural Language Processing (NLP) expert educated in Taiwan and further in Germany in the discipline of computational linguistics and NLP. With a professional career built on NLP-related research and development in leading companies such as Philips and Nuance, he later founded an NLP-led company Lingtelli, which Riversoft acquired, for whom Claude now works as the Chief Technology Officer. 

Riversoft is a Taiwan-based software development company that specializes in generative A.I. chatbots and Natural Language Processing (NLP) technology. They serve a broad client base spanning from Travel, Healthcare to E-commerce. Today, we are privileged to have Claude sharing his insights on the A.I.-dominant era and the core value of Riversoft.

Current Status and Prospects of the A.I. market

A.I. market is predicted to experience robust growth in the next decade. According to Next Move Strategy Consulting, its current value of nearly 100 billion U.S. dollars is projected to increase twentyfold by 2030, reaching almost 2 trillion U.S. dollars.

Before the advent of ChatGPT, notable advancements in A.I. were already observed in specific fields. For instance, AlphaGo, an A.I. software, achieved remarkable success by defeating top-level players in the game of Go with minimal learning time. The emergence of ubiquitous technologies like ChatGPT has transformed A.I. into a tool that permeates every aspect of our lives. OpenAI, the renowned company behind ChatGPT and GPT4.0, is a major player in this field. Anthropic, founded by former OpenAI engineers, is its primary competitor. Anthropic recently announced that they had expanded their chatbot’s context window to 75,000 words. To put it into perspective, their chatbot can process an entire novel in just one minute, while a human would require approximately five hours to do the same.

These generic models can now be readily applied across various industries by anyone. Students can utilize them for homework assistance, doctors can employ them to analyze clinical notes, and companies like Riversoft can leverage their presence to enhance customer service.

According to Claude, we can anticipate industry giants focusing on developing cutting-edge and general A.I. models. At the same time, downstream companies will build upon these foundations, tailoring and applying them to specific domains.

“It takes a huge amount of resources to create and build those models. So it’s not for small and middle-sized companies. We have to think about (is that) now we have ChatGPT and GPT-like models, what can we do more to go deeper into our domains? We have to know our real value instead of just using ChatGPT. This is a huge opportunity.”  Claude explains.

Natural Language Processing (NLP) & Generative A.I.

Claude explains that NLP is a broad term encompassing various technologies and applications. The transformer model is one of the NLP models with inherent features making it particularly well-suited for processing language data. The “GPT” in ChatGPT stands for “Generative Pre-trained Transformer.” The neural network structure of a transformer draws inspiration from the intricacies of the human brain, which may explain why interacting with transformer-based models feels akin to engaging with humans in conversation.

“Everything converges down to a Transformer based architecture to tackle different kinds of problems, especially now that we have ChatGPT and GPT-like things that seem to solve every aspect of NLP. “

Riversoft’s Edge and Core Values

“ChatGPT and similar technologies can significantly enhance our ability to serve customers, but rather than relying solely on them, we need to identify our values and advantages,” emphasizes Claude.

In Claude’s view, the essence of research and application in NLP technology lies in understanding meaning. People desire machines to communicate, read, and write like humans, and this hinges on comprehension. Although GPT and similar models can generate articles that appear human-like, the extent to which they truly understand their output is still debatable.

“ChatGPT and similar models are super capable of generating human-like patterns, but we still cannot say that they understand what they produce. Maybe it’s a philosophical issue; I mean sometimes you think you have free will you think you have consciousness but some would argue that it’s illusion.”

Presently, one of Riversoft’s products in the medical field involves analyzing clinical notes written by doctors, extracting relevant information, and presenting it in a structured manner.

Riversoft’s goal is to delve deeper into information generated by A.I. models and humans. While such information is easily comprehensible for humans, it is not readily accessible to computers. By developing a system capable of processing unstructured data for further use, Riversoft aims to offer highly detailed and personalized travel itineraries by scouring travel sites and blogs. In the medical realm, the system can analyze medical records, alerting healthcare professionals to any cautionary factors when prescribing treatments and medications. This system streamlines necessary checks and reduces the risk of inappropriate treatments.

“We want to generate a core that can process through all the content and put it back to the system where the system can orchestrate all the information, be it structured or unstructured, and put them all into a very systematic process. So that we can serve our client,” states Claude.

Consciousness of A.I.

“If you don’t have the knowledge to verify if what’s been generated (by A.I. models) is hallucination or facts, you will be misinformed. I think we provide here that no matter what kind of generation models, we will have a way to go through the content and modify and deletions.” “It’s dangerous to rely only on one source. So it’s important to have a mechanism to correct and modify the output of those models.”

The technology employed by Riversoft strives to increase solution autonomy and flexibility. Through a combination of deep learning models and semantic mechanisms, human intervention can be minimized within the system. This is how small and medium-sized companies can thrive in the era of the A.I. revolution.

Here to help or here to replace? Possible changes that A.I. might bring about

The disruptive breakthrough of A.I. has sparked diverse perspectives among scholars and researchers. While some express optimism, others raise concerns about its implications. Geoffrey Hinton, widely regarded as the godfather of A.I., even resigned from Google to adopt a more neutral stance on the matter. He shares several major concerns.

Firstly, Hinton highlights the issue of misinformation. With programs capable of generating words, pictures, and videos, discerning reality from fiction on the internet will become increasingly challenging. Secondly, he emphasizes the impact on the job market. While A.I. currently complements human work, there is growing apprehension that as A.I. capabilities expand, it will disrupt the labour market on a large scale. Lastly, although still hypothetical, Hinton worries about the existential risk that may arise when all human work is replaced by A.I.

In contrast, Claude adopts a more optimistic view of the issue, focusing on the practical implementation of these technologies. “It’s less about controlling the research and more about how we put these technologies to use,” he states.

While the ability to generate information is currently in the spotlight, Claude emphasizes the importance of appropriateness. Generated information can be either factual or mere fabrications. Without domain-specific knowledge, distinguishing between the two can be challenging. This underscores the need for mechanisms to modify and correct the output of A.I. models. Humans possess ethical, legal, and societal constraints that guide us in distinguishing between good and evil. Similarly, it is crucial to establish similar systems to exercise control over A.I.

Claude reassures, “We are not yet at the stage of sci-fi scenarios where machines are trying to eradicate the human race.” However, we can already witness A.I. making significant changes in various fields. For instance, as autonomous driving technology matures, human driving—the source of many uncertainties—can be eliminated, making roads safer for everyone.

Since we are still far from a world where machines entirely replace humans, worrying about such scenarios is unrealistic. Instead, Claude suggests that we prepare society for a future where A.I. plays an increasingly prominent role in our lives.

Regarding the impact on the job market, Claude offers a more positive perspective. “Currently, many people perform mundane, repetitive work not because they want to, but out of necessity.” He envisions a society where labor-intensive jobs are replaced by machines. Claude proposes that as firms’ personnel costs decrease due to A.I., the resulting excess profit could be redistributed by the government through initiatives like universal basic income—an idea advocating for regular cash transfers to citizens without conditions.

Furthermore, with A.I. taking over tasks such as handling emails and questionnaires, individuals will have more time and resources at their disposal. They can spend quality time with family and friends or just look at the stars. “A.I. can make humans more human by liberating them from industrialized processes,” Claude asserts.

Overall, while acknowledging the potential changes brought about by A.I., Claude maintains an optimistic outlook, emphasizing the importance of responsible implementation and preparing society for a future where A.I. increasingly shapes our lives.

How ChatGPT-like Models Improve Collaboration in Workplace  

ChatGPT-like models have the potential to enhance collaboration in the workplace significantly. Their consistent and reliable nature eliminates fluctuations and forgetfulness that humans may experience. This is reflected in the significant increase of 44% in the use of collaboration tools between 2019 and the present, as reported by Gartner. While collaboration is becoming increasingly prevalent, managing and tracking teamwork can be demanding and time-consuming.

With systems like ChatGPT, all communication and organization needs can be efficiently coordinated and consistently followed. “You can imagine having an omnipotent assistant standing next to you, telling you things to beware of, to take care of, and making sure everything stays on track,” said Claude.

Moreover, ChatGPT and similar systems act as connectors within teams. Previously, companies relied on fixed systems that required strict adherence to established procedures and disciplined execution. However, ChatGPT-like models function as proactive agents, identifying missing elements and contacting individuals to ensure tasks are completed. “The old system required rigid disciplines from everyone. But sometimes it doesn’t work because people failed to follow up.” This new system effectively coordinates and controls the workflow, automating various office processes.

Importantly, the applications of such models extend beyond the workspace. Claude gave a scenario when planning a trip to London; it is essential to confirm the availability of desired attractions and restaurants and check the weather and traffic conditions. Here, an agent working in conjunction with other models can simplify and streamline the process for tourists, providing coordinated assistance and relevant information.

The Challenges

When discussing the challenges faced by Riversoft, Claude acknowledged the intense competition within the A.I. industry as a significant hurdle. Competitors are constantly striving to outperform and replace each other. However, he expressed confidence in Riversoft’s technological advantage in deep learning models and semantic technology, which, when combined, serve as their competitive edge.

Financial and market-related challenges are also prevalent, including the need to attract investments and establish a presence in the market. The impact of the pandemic has led to a slow recovery of capital, making it more difficult to secure the necessary funding. Additionally, as these technologies are relatively new, there are no established models to follow, necessitating significant time and effort to communicate with customers and educate them about the benefits. Acquiring customers in this landscape can be costly, and Riversoft aims to address this challenge by exploring strategies to reduce customer acquisition costs.

Currently, Riversoft’s client base extends across Taiwan, Japan, Singapore and Europe. Their strategy involves partnering with major players in different domains, enabling them to gain valuable insights and identify the essential features required in each market. As the technology continues to mature, Riversoft intends to expand its market share and explore new opportunities.

Keeping up with the Trend 

As the CTO of Riversoft, Claude emphasized the importance of staying updated with the latest trends and advancements in the field. Platforms like YouTube and Google serve as valuable sources of general information, allowing individuals to discover experts who provide insights into specific areas. This enables them to keep abreast of the latest developments. While academic papers are a valuable source of information, Claude acknowledged their limitations in terms of speed and relevance to the market. “I spend lots of time keeping myself updated. In such an ever-changing field, if you don’t survey for a week, you find yourself missing out on important things,” said Claude.

Embracing Innovation, Collaboration, and Endless Possibilities 

ChatGPT and similar A.I. technologies have brought about a transformative change in our daily lives. From the astonishing growth of the A.I. market to the intense competition among industry players, it is clear that A.I. is here to stay and will continue to shape the future.

While the challenges of financials, market-entry, and customer acquisition persist, companies like Riversoft leverage their technology edge and strategic partnerships to carve a niche in specific domains such as travel and medicine.

As we embrace this A.I.-prevalent era, it becomes crucial for individuals and organizations to adapt and make the best use of A.I. tools. “The most important aspect of human intelligence is creativity,” said Claude. By understanding the potential of A.I. and harnessing its capabilities, we may be able to navigate the evolving landscape and unlock endless new possibilities for innovation, collaboration and beyond.

References

• https://www.theverge.com/2023/5/12/23720964/chatbot-language-model-context-window-memory-anthropic-claude
• https://www.statista.com/statistics/1379027/chatgpt-use-us-companies-money- saved/
• https://lilianweng.github.io/lil-log/assets/images/transformer.png
• https://www.statista.com/statistics/1365145/artificial-intelligence-market-size/
• https://www.nytimes.com/2023/05/01/technology/ai-google-chatbot-engineer- quits-hinton.html