Caithness Map :: Links to Site Map Great value Unlimited Broadband from an award winning provider TUE 15TH JUL 2025    3:36:01 AM BST
This site uses cookies, by continuing to use this site you accept the terms of our privacy policy
Back To Top
Caithness.Org Quick Links
Home
Construction
Leisure
Manufacturing
Misc.
Primary
Professional
Public
Retail
Tourism
Transport
Site Map
 
 
 
 
 
 
 
 
 
 
 
 
 

Feed 2.0 Loading...

The New Kid on the AI Block 'Deep Seek' Is Upsetting the Older Players

28th January 2025

The latest AI chat is from a Chinese company and its name is DEEP SEEK.

Wall Street had a shock to the system on concerns that a cheap artificial intelligence-model from Chinese startup DeepSeek could make valuations that have powered the bull market tough to justify.

From New York to London and Tokyo, equities got hit hard on Monday. And while the slide did come after a torrid rally, the underlying reasons behind Monday's rout could cause longer-term damage.

The Nasdaq 100 sank 3% while the S&P 500 dropped 1.5%. A closely watched gauge of chipmakers also plunged the most since March 2020. Nvidia, sank 17%—the biggest market-cap loss for a single stock ever - wiping $593 billion from its market value.

On Monday it cause a massive drop in the share price of Nvidia the biggest and richest chip company quoted on the stock markets.

So here is a brief outline of AI and the chatbots.

The cost of operating an AI model like DeepSeek, ChatGPT, or Microsoft Copilot depends on several factors, including the scale of deployment, the complexity of the model, infrastructure efficiency, and the specific use cases. DeepSeek has apparently done the work with much cheaper chips.

1. Model Size and Complexity
DeepSeek: If DeepSeek is a smaller or more specialized model compared to ChatGPT or Microsoft Copilot, it could be cheaper to operate. Smaller models require less computational power and memory, reducing infrastructure costs.

ChatGPT (OpenAI): OpenAI's GPT-4 and other large models are computationally expensive to run due to their size and complexity. They require significant GPU/TPU resources, which increases operational costs.

Microsoft Copilot: Copilot is built on OpenAI's GPT models but is integrated into Microsoft's ecosystem (e.g., GitHub, Office). While Microsoft benefits from economies of scale and optimized infrastructure, the underlying model costs are still high.

2. Infrastructure and Optimization
DeepSeek: If DeepSeek is designed with efficiency in mind (e.g., using quantization, distillation, or other optimization techniques), it could reduce operational costs.

ChatGPT: OpenAI has invested heavily in optimizing its models for inference, but the sheer size of GPT-4 means it remains expensive to operate at scale.

Microsoft Copilot: Microsoft leverages its Azure cloud infrastructure, which is highly optimized for AI workloads. However, the integration of Copilot across multiple products (e.g., Teams, Word, Excel) increases the overall operational complexity and cost.

3. Scale of Deployment
DeepSeek: If DeepSeek is deployed for niche or specific use cases with lower user demand, its operational costs would likely be lower than those of widely used systems like ChatGPT or Copilot.

ChatGPT: ChatGPT has a massive user base, requiring significant infrastructure to handle millions of queries daily. This scale drives up costs.

Microsoft Copilot: Copilot is integrated into widely used Microsoft products, meaning it operates at an even larger scale than ChatGPT, further increasing costs.

4. Use Case and Customization
DeepSeek: If DeepSeek is tailored for specific tasks (e.g., customer support, data analysis), it might require fewer resources than general-purpose models like ChatGPT or Copilot.

ChatGPT and Copilot: These models are designed for general-purpose use, requiring more computational resources to handle a wide variety of tasks and queries.

5. Development and Maintenance Costs
DeepSeek: If DeepSeek is a newer or less complex system, its development and maintenance costs might be lower compared to the extensive R&D and ongoing updates required for ChatGPT and Copilot.

ChatGPT and Copilot: Both systems require continuous updates, fine-tuning, and monitoring, which add to operational expenses.

6. Energy Consumption
Larger models like GPT-4 and Copilot consume significantly more energy during training and inference, contributing to higher operational costs. A smaller or more efficient model like DeepSeek could have a lower energy footprint.

Conclusion:
While it's difficult to provide a definitive answer without specific cost data, DeepSeek is likely cheaper to operate than ChatGPT or Microsoft Copilot if it is a smaller, more specialized, and optimized model. However, if DeepSeek scales up to handle similar levels of demand or complexity, its costs could increase accordingly. The operational cost of any AI system ultimately depends on its design, deployment scale, and efficiency optimizations.

If you have not tried them so far here are some to get you started
https://chat.deepseek.com/
https://copilot.microsoft.com/
https://chatopen.app/onboarding

More ---
Chat models, like other AI technologies, can generate revenue and earn a profit through various business models and strategies. Here are some of the most common ways chat models can be monetized:

1. Subscription Services
Freemium Model: Offer basic features for free and charge users for premium features, such as advanced capabilities, faster response times, or access to specialized knowledge.

Enterprise Subscriptions: Provide businesses with tailored solutions, such as customer support automation, data analysis, or internal knowledge management, for a recurring fee.

2. Pay-Per-Use or API Access
API Licensing: Charge developers and businesses for API access to integrate the chat model into their own applications, products, or services. Pricing can be based on the number of API calls, tokens processed, or response volume.

Usage-Based Pricing: Allow users to pay only for what they use, which can attract smaller businesses or individual developers who don't need a full subscription.

3. Advertising and Sponsorships
In-Chat Ads: Display relevant advertisements within the chat interface, similar to how search engines or social media platforms monetize their services.

Sponsored Content: Partner with brands to integrate sponsored responses or recommendations into the chat model's outputs.

4. B2B Solutions
Customer Support Automation: Sell chat models as a tool for automating customer service, reducing the need for human agents and improving efficiency.

Internal Knowledge Management: Offer chat models as a way for businesses to streamline internal communication, document retrieval, and employee training.

Industry-Specific Solutions: Develop specialized chat models for industries like healthcare, finance, or legal services, where tailored AI solutions can command higher prices.

5. Data Insights and Analytics
Data Monetization: Analyze anonymized chat data to provide insights to businesses or researchers. For example, a chat model could identify trends in customer inquiries or market sentiment.

Consulting Services: Offer consulting services based on the insights generated by the chat model, helping businesses optimize their operations or marketing strategies.

6. Partnerships and Integrations
Platform Integrations: Partner with existing platforms (e.g., Microsoft Teams, Slack, or Shopify) to integrate the chat model as a value-added feature, sharing revenue with the platform provider.

Co-Branding: Collaborate with other companies to create co-branded solutions, such as AI-powered virtual assistants for specific products or services.

7. Content Creation and Curation
Content Generation: Charge for generating high-quality content, such as blog posts, marketing copy, or social media updates, using the chat model.

Personalized Recommendations: Offer personalized content recommendations (e.g., books, movies, or products) and earn affiliate revenue or commissions.

8. Education and Training
E-Learning Platforms: Provide AI-powered tutoring or language learning services, charging users or institutions for access.

Corporate Training: Develop chat-based training programs for businesses, helping employees learn new skills or comply with regulations.

9. Gaming and Entertainment
Interactive Storytelling: Create immersive, AI-driven storytelling experiences for gamers or readers, charging for access or in-app purchases.

Virtual Companions: Offer virtual companions or chatbots for entertainment purposes, with premium features available for purchase.

10. Research and Development
Collaborations with Academia: Partner with universities or research institutions to develop new applications for chat models, funded by grants or joint ventures.

Custom AI Solutions: Develop custom chat models for specific research projects or industries, charging for the development and deployment of these solutions.

11. Licensing and White-Label Solutions
White-Label Products: License the chat model to other companies, allowing them to rebrand and sell it as their own product.

Franchise Models: Offer franchises or regional licenses for businesses to operate and monetize the chat model in specific markets.

12. Crowdsourcing and Community Contributions
User Contributions: Allow users to contribute data or improvements to the chat model in exchange for credits, discounts, or access to premium features.

Open Source with Paid Support: Offer the chat model as open-source software but charge for professional support, customization, or hosting services.

Challenges to Profitability
While there are many ways to monetize chat models, there are also challenges:

High Operational Costs: Running large-scale AI models requires significant computational resources, which can eat into profits.

Competition: The market for AI chat models is becoming increasingly crowded, driving down prices and margins.

Ethical Concerns: Monetizing user data or using ads in chat models can raise privacy and ethical issues, potentially alienating users.

Conclusion:
Chat models can earn a profit through a combination of subscription services, API access, advertising, B2B solutions, and innovative use cases. The key to profitability lies in identifying the right target audience, optimizing operational costs, and continuously improving the model to deliver value. As the technology evolves, new monetization opportunities are likely to emerge, making chat models a versatile and potentially lucrative tool for businesses.


Even more
As of October 2023, there are hundreds, if not thousands, of chat models already online, ranging from large-scale, general-purpose models like OpenAI's ChatGPT to specialized, niche models designed for specific industries or tasks. The exact number is difficult to pin down due to the rapid proliferation of AI technologies and the fact that many models are privately deployed by companies, researchers, and developers. Below is a breakdown of the landscape:

1. Major General-Purpose Chat Models
These are the most well-known and widely used chat models, often developed by large tech companies or AI research organizations:

OpenAI's ChatGPT (GPT-3.5 and GPT-4): One of the most popular and advanced chat models, used by millions of users and integrated into various applications.

Google's Bard: A conversational AI model built on Google's PaLM 2 architecture, designed to compete with ChatGPT.

Microsoft's Copilot: Powered by OpenAI's GPT models, integrated into Microsoft products like Bing, Office, and GitHub.

Anthropic's Claude: A competitor to ChatGPT, focused on safety and ethical AI.

Meta's LLaMA: A family of open-source models that can be fine-tuned for chat applications.

Cohere's Command: A chat model designed for enterprise use cases, such as customer support and content generation.

2. Open-Source Chat Models
Open-source models have proliferated, allowing developers to deploy their own chat models for free or at a lower cost:

LLaMA (Meta): A foundational model that has been fine-tuned by the community for chat applications.

Falcon (by TII): A high-performance open-source model used for chat and other NLP tasks.

Vicuna: A fine-tuned version of LLaMA, optimized for conversational AI.

Alpaca: A lightweight, open-source model based on LLaMA, designed for research and small-scale applications.

Mistral 7B: A compact and efficient open-source model for chat and other tasks.

3. Industry-Specific Chat Models
Many companies and organizations have developed chat models tailored to specific industries or use cases:

Healthcare: Chat models for medical diagnosis, patient support, and mental health counseling.

Finance: AI chatbots for customer service, fraud detection, and financial advice.

Legal: Chat models for contract analysis, legal research, and client interaction.

Education: AI tutors and virtual teaching assistants.

Retail and E-commerce: Chatbots for customer support, product recommendations, and order tracking.

4. Regional and Language-Specific Models
Chat models are being developed to cater to specific languages and regions:

China: Baidu's ERNIE Bot, Alibaba's Tongyi Qianwen, and Tencent's Hunyuan.

India: AI models tailored for Indian languages and cultural contexts.

Europe: Multilingual models like DeepSeek and others designed for European languages.

5. Custom and Proprietary Models
Many companies and startups have developed their own chat models for internal use or as part of their products:

Customer Support Chatbots: Deployed by companies across industries to handle inquiries and support tickets.

Virtual Assistants: Integrated into apps, websites, and devices to provide personalized assistance.

Gaming and Entertainment: Chat models used in interactive storytelling, virtual companions, and gaming.

6. Research and Experimental Models
Academic institutions and research organizations have developed numerous chat models for experimentation and innovation:

Stanford's Alpaca: A research-focused model based on LLaMA.

EleutherAI's GPT-NeoX: An open-source model used for research and development.

Hugging Face's Models: Thousands of fine-tuned chat models available on the Hugging Face Hub.

7. Small-Scale and Niche Models
Smaller companies and individual developers have created chat models for specific tasks or audiences:

Personalized Chatbots: For hobbies, niche interests, or community engagement.

Localized Solutions: Chat models designed for small businesses or local communities.

Estimated Numbers
While it's impossible to provide an exact count, here's a rough estimate:

General-Purpose Models: Dozens (e.g., ChatGPT, Bard, Claude, etc.).

Open-Source Models: Hundreds (e.g., LLaMA, Falcon, Vicuna, etc.).

Industry-Specific Models: Thousands (deployed by companies worldwide).

Custom and Proprietary Models: Tens of thousands (used by businesses and developers).

Growth Trends
The number of chat models online is growing rapidly due to:

Advancements in AI Research: Easier access to pre-trained models and fine-tuning tools.

Open-Source Movement: Lowering the barrier to entry for developers.

Cloud Computing: Making it easier to deploy and scale chat models.

Demand for AI Solutions: Increasing adoption across industries.

Conclusion
There are already hundreds to thousands of chat models online, with more being developed and deployed every day. The landscape is diverse, ranging from massive general-purpose models to niche, specialized solutions. As AI technology continues to evolve, the number of chat models is expected to grow exponentially, driven by innovation, accessibility, and demand.

Who is behind DeepSeek

Chinas DeepSeek AI shakes industry and dents Americas swagger

DeepSeek AI launch sees a trillion dollars wiped off world's biggest tech companies' share prices