As one of the few teams and companies with the capacity to build LLMs with hundreds-billion-parameters, at Stellaris AI, our target is to bring native secured, reliable AI to the world.
Since the end of last year, ChatGPT has induced a revolution across AI area and every industries. And the core of this AI revolution is that the LLM, contains hundreds-billion-parameters and it has a so called emerging ability, which means that it can complete tasks it has never been trained on before. At Stellaris AI, we stick to the principle of building such a general purpose foundation model with hundred billion parameters.
On May 10th, we have successfully released SGPT-2.5, which is the first and only proprietary large language model with hundreds of billions of parameters in Hong Kong. We have tested our models comprehensively across a range of benchmarks which all give a impressive score, compared to other main-stream commercial models.
According to the benchmark tests, our model has a performance between GPT-3.5 and GPT-4. The anticipated timeline of SGPT for reaching the technical capabilities of GPT-4 is estimated to be within six months to one year
Besides that, our models has very unique technological strength and features, which differentiate us from other large language models.
Firstly, our model has a high level of security protection, it can easily deal with security breaches which bother other LLMs like Jailbreaking.
Secondly, our model has a very good real time learning ability. It always deliver to the users the most up-to-date answers. And also we employ external knowledge references to reduce the level of hallucinations a lot.
Thirdly, we apply a mode of local deployment for our users, which means our users can run our models internally on their servers and this frees them from risk of data security, IP and compliance concerns.
To evaluate the performance of the SGPT model, three key benchmarks were used: the MMLU dataset (comprehensive ability testing), the HellaSwag dataset (logical reasoning and continuity), and the ARC dataset (complex professional knowledge). The SGPT model excelled in these tests, outperforming other popular models such as OpenAI GPT-4, Anthropic Claude, and those from major companies like Meta and Google. Notably, the SGPT-2.5 variant consistently ranks in the top three positions across these evaluations. This stellar performance highlights the model's practical capabilities and signifies the company's commitment to delivering top-tier value to its stakeholders.
This summarizes the 3 investment highlights for our companies.
1st, we have already developed and released a first-class, Top-tier, hundred-billion-parameters LLMs. Stellaris AI's SGPT is a pioneering, top-tier LLM with a hundred billion parameters, setting industry benchmarks. Its features, highlighting practical knowledge assimilation, redefine AI standards and versatility. We maintain exclusive IP rights, safeguarding and refining our innovations. Despite limitations, Stellaris AI's resource optimization consistently produces advanced AI models, underscoring our unwavering dedication to AI innovation.
Secondly, Stellaris AI takes pride in the formation of a multifaceted and highly competent team of eight core members, all are PhDs in Computer Science and AI. Each individual in our team brings a unique set of skills and expertise to the table, enabling us to deliver unparalleled value in the Artificial Intelligence domain, particularly in Large Model Training, High Performance Computing, and AI Applications and Commercialization.
Thirdly, Stellaris AI is headquartered in Ireland, leveraging the country's prominence in the tech sector and its favorable business environment for global growth. The company is aggressively expanding, focusing on key global markets including the APAC and EMEA regions. Beyond expansion, Stellaris AI is forming strategic alliances not just with commercial entities, but also with educational institutions worldwide. These collaborations broaden their network and open up avenues for co-innovation. With its strategic location, global outreach, and diverse partnerships, Stellaris AI positions itself as a growing player in the Large Language Models sector and invites investors to capitalize on its unique market stance.
Lastly, we have pioneered a natively secured SGPT model, ensuring superior security from design inception. This stands in contrast to patch-based methods used by others like OpenAI. Stellaris shift the focus from patched model security to native model safety, a transformative methodology change. Deep-rooted safety protocols are embedded at foundational levels, offering comprehensive protection. This contrasts with OpenAI's surface-level, reactive solutions. Stellaris ensures defense against varied threats, including privacy breaches and model violations. We've also addressed the "triangle curse" of model training by accessing expansive multi-sourced data securely, enabling growth beyond traditional limitations.
Stellaris AI's business model is centered on its unique SGPT technology, forming an expansive ecosystem. We function on two main pillars: Model Services and Computational Power Services. Initiatives such as the Stellaris Partners Program and Stellaris A/H100 GPT Farms enhance collaboration and optimize resources. Stellaris offers both software and hardware solutions, aiming for wide industry relevance. Our ultimate goal is to be a global infrastructure leader, leveraging their SGPT tech and strategic approach.
Stellaris AI's Model Services, inspired by Apple's developer program, utilize SGPT the foundation model and Stellaris A/H100 GPU Farms to deliver a range of services. Our revenue approach promotes shared growth, with both clients and Stellaris profiting as client services expand. We provides AI tools to both B2B and B2C clients, making technology more accessible and affordable. We emphasize strong client relations by regularly updating their models and offering consistent support, aspiring to be central to all AI-enhanced user products.
Stellaris AI differs from competitors by centralizing its computational resources, bypassing expensive public cloud fees. We prioritize LLM development, model security, and robust computational infrastructure. By managing their A/H100 GPU farms, We ensure data security, privacy, and affordability. We've redefined computation as a revenue source rather than just an expense. Our unique profit model emphasizes both model services and computational power, setting them apart from other LLM providers.
Overall, Stellaris AI is positioned for growth and leadership in the LLM field and seeks investor participation in our innovative journey.
ความคิดเห็น