Alivever.com |
About | Projects | Career | Blog |
Introduction |
||||||||||||||||||||||||||||||
We are a software and AI research team from Berlin, Germany.
Its core members have 10-25 years of work experience, including: large-scale internet search engine, large-scale real-time trading system, large-scale data center, AI chips and AI algorithm. |
||||||||||||||||||||||||||||||
Tech and Route |
||||||||||||||||||||||||||||||
1. We will attempt to develop an active-learning, self-evolving, focused, Small-scale, General AI pototype. a. Active-learning: It will actively ask AGI, API, and humans for the knowledge it want, rather than starting from pure text in passive original training. b. Self-evolving: Its can make learning plans for itself, and even codes for itselfs next version. Engineers only provide the final goal and acceptance criteria. c. Focused: It will focus only on specific domains. Once it achieves the learning goal, it‘ll stops and be seen as grown up, rather than continuous acquiring all other information like a black hole. d. Small-scale: Each AI can run independently on a gaming GPU. That means the maximum required GPU memory less then 24GB. e. General AI pototype: Its initial embryo and evolutionary method are universal. But finally we get different AIs for different goals. 2. Diversified AI network. a. We hope to cultivate various AIs to be our employees, and provide personalized services for users. c. AI Network: Multiple small AIs can collaborate to discuss and solve complex problems. |
||||||||||||||||||||||||||||||
Mission |
||||||||||||||||||||||||||||||
1. To contribute necessary data services and technological explorations for promoting "AI diversification" and "explainability and controllability of AI computing processes." a. Data Services.We do not provide traditionally directly downloadable text and multimedia datasets, but rather deeply process publicly available datasets to ultimately offer data services that can be directly utilized by AI training and computation. b. On "AI explainability and controllability": While single large models are effective, their drawbacks are evident: the thinking process is unexplainable and uncontrollable. This ultimately leads to risks of uncontrolled outcomes. We hope to explore an AI structure different from purely parametric models to address this issue. 2. To explore technological and ethical aspects in promoting the recording and inheritance of individual human experiences. a. Individual-oriented: We focus on each specific person rather than the overall human civilization processes (history, economy, religion, science, etc.). We model and record based on their lifetime experiences, not limited to their activities on the internet (clicks, comments, purchases, views, etc.). b. Technology and Ethics: We lean towards integrating the latest AI technologies from an applied perspective rather than academic research papers. Additionally, we will pay significant attention to ethical issues that these AI technologies may touch upon in their applications. We believe that technology ethics is not a stumbling block to AI development; rather, it is a guide steering AI research in the right direction. |
||||||||||||||||||||||||||||||
Vision |
||||||||||||||||||||||||||||||
Assist in establishing a diversified AI network globally, ensuring its operation is interpretable and human-controllable. |
||||||||||||||||||||||||||||||
Donations |
||||||||||||||||||||||||||||||
We are seeking a small amount of funding for purchasing hardware required for computation and covering basic expenses for volunteers and interns. We will publicly disclose the usage of donations from each contributor to ensure transparency in fund utilization. We hope to receive donations for the following:
If you are willing to contribute, we will provide the following benefits to you: |