Genspark’s Vision of the ‘Next-Generation AI’: From Models to Services
IT DAILY ·
✦ AI Summary
Competition in the generative AI market is shifting from model performance to services that connect to real work outcomes.
Conventional generative AI is said to have limits in understanding the user’s entire workflow and providing ongoing support.
Genspark has proposed a shift from a model-centered approach to a service-centered one, along with its “SecondBrain” and Super Agent strategies.
Competition in the generative AI market is shifting from performance battles in LLM training and inference to service competition that links directly to real work outcomes. As the performance of generative AI models such as OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude has improved rapidly, AI use cases have expanded to include document writing, information retrieval, coding, and image generation.
However, conventional generative AI is seen as having limitations, including a lack of understanding of the user’s entire workflow and insufficient ability to provide ongoing support. In other words, even as increasingly capable models are being used for a wide range of tasks, constraints remain when it comes to understanding the full work process and continuing to assist throughout it.
Genspark, a Silicon Valley-based AI startup founded in the U.S., sees these limitations as the core challenge in the next-generation AI services market. Genspark believes that developing models that simply produce better answers is not enough, and that AI’s ability to remember user context, make judgments, and complete tasks end to end will become the key to future AI service competitiveness.
Genspark has presented a shift from a model-centered approach to a service-centered one as the direction for next-generation AI. Its point is that services matter more than models in next-generation AI. One image example it cited was a pin cherry flower against a concrete wall background, and the image quality was evaluated as comparable to a high-resolution photograph.
Genspark started out as an AI search engine and is now described as a frontier AI player standing out in image generation and PPT creation. This description is a summary of its solutions and services based on Genspark’s website and foreign media reports from Forbes, Nikkei, CNBC, and others.
Genspark was founded in Palo Alto in December 2023. Palo Alto is at the heart of Silicon Valley, where Stanford University is located, and has infrastructure concentrated with cutting-edge AI models, semiconductors, capital, and talent. Genspark is rolling out its services by working with frontier AI labs and rapidly applying the latest AI technologies to its products.
CTO Cai Zhu said AI models’ capabilities have improved dramatically compared with 1 or 2 years ago. At the same time, he said it is difficult for ordinary users to accurately assess differences in performance between models. This suggests that competition in the AI market may move beyond simple comparisons of model performance.
Models themselves are improving rapidly. But the development of user experience and service structures that connect them naturally to everyday work is still presented as an area that remains underdeveloped. As a result, it is also being highlighted that the focus of AI competition may shift not only to model performance but also to the experience and service structures through which users actually connect AI to their work.
Genspark’s founders are focusing not on a lack of AI intelligence, but on the complexity of work in the process of using AI. COO Wen Shan said the reason information is needed is not knowledge acquisition itself, but completing work. Accordingly, if the key outcome in the search era was finding the necessary information, in the generative AI era the steps after information retrieval are presented as the more important point. The scope of AI support also extends beyond information acquisition to judgment, drafting, sharing, and follow-up work.
Founder and CEO Eric Jin likened current AI to a genius with the memory of a chicken. He said AI’s abilities are outstanding, but in multiple conversations it forgets existing premises and priorities. As a result, users have to repeatedly explain identity, work details, and priorities.
Genspark believes that AI’s lack of contextual retention is one reason work-related hassles have not been fully eliminated even after the spread of generative AI. It therefore sees generative AI alone as unable to sustain work context, leaving inconveniences in place.
Based on this view, Genspark treats AI model development and AI service development as separate areas and distinguishes its business from AI model developers. It uses the analogy that while OpenAI and Anthropic make engines, Genspark makes the car that uses that engine to reach the destination. Genspark also focuses on “what work” rather than “what model.”
This reflects the judgment that user demand lies less in comparing AI models and more in organizing materials, replying to emails, preparing meetings, and advancing projects. With that in mind, Genspark presents the concept of “SecondBrain” as a solution.
SecondBrain is a continuous brain for knowledge workers, and it treats workflow information such as email, meetings, chats, documents, applications, and project histories as links to be connected. Its purpose is to connect diverse information into a single context, enabling AI to maintain a continuous understanding of the user’s work.
Genspark’s latest service, “Genspark 6.0,” consists of 4 layers: SecondBrain, Super Agent, Build Suite, Office Suite, Content Suite, and ZenTeam. The core, however, is explained as connecting a single workflow of memory, judgment, generation, and collaboration rather than the individual functions of the 4 layers. This also aligns with the point that Genspark’s AI service strategy is notable for avoiding an approach that confines AI within a software screen.
Genspark places the meeting and conversation record itself outside the purpose, and aims to automatically carry out follow-up work based on those records. In other words, it presents AI not as a simple recording tool, but as a system that handles subsequent tasks based on the record.
The goal of this direction is to expand the scope of “context” that AI handles from digital data to real-world work activities. Information to be understood includes non-fully digitized data such as conversations in meeting rooms, notes taken on the move, and face-to-face meetings. The company said it is not only about access to the latest AI models, but also about giving AI access to the physical world.
The competitive criteria for next-generation AI services are expanding beyond answer accuracy to include the breadth of context captured and the ability to continuously support long-term work processes. The article explains that the competitive factors for next-generation AI services are shifting from simply generating correct answers to securing broader context and providing sustained work support.
In line with this shift, Genspark has presented “SecondBrain” and Super Agent strategies to respond to market changes. The direction of next-generation AI development lies beyond response tools in chat windows and toward remembering user context, making judgments, and continuously carrying out real work, and the key to winning or losing in the next-generation AI services market is also said to be AI’s ability to remember user context, make judgments, and continuously carry out real work.
Source: IT DAILY · Jo Min-soo
Original: https://www.itdaily.kr/news/articleView.html?idxno=241448
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.