Deepseek | Deep Thoughts | Deep Insights

Mohamed Haneef

What you seek is seeking you - Rumi

?The quote "What you seek is seeking you" is attributed to the 13th-century Persian poet and philosopher Rumi.

It's one of his most famous lines, often interpreted as a reminder that the things we desire or are searching for in life are, in some sense, already drawn toward us as well.

?I am sure you will sense the same thing with the output that you derive from the tech-stacks as well…In the sense that the tools/software provides only the output that you are expecting from it, because it knows you of 'what you really want ?'. The other way to say,

the devices like the phones or tablets or the apps that you use knows you better than your friends, parents or your children.

?Sometimes it provides you a better insights of yourself….

?Since everyone was talking about 'Deepseek' in last few days, I thought let me also try it out. After a bit of login issues, I was able to login successfully with my 'Apple Identity' since I downloaded it from the Apple Appstore.

?Then, I asked the same questions to?3 of my tech-friends (chat-bots)…I asked a simple question to our Chat-bot friends: "Who am I?"

Comparison

Firstly, to my new friend 'Deepseek'. A friend who shook the world by creating the huge financial impact due to its surprising entry last week. The overall lose of shares value by almost 2 Trillions in which NVIDIA alone has lost ?nearly $600 billion in one session, the largest single-day drop in market history.

?Okay…Now, what is the response?

The response is no-where matching my expectation. Because, it is just a new software and it has not interacted with me well enough yet to reflect back to me on 'Who Am I ?'

Secondly, I did ask the same question to Gemini and the Copilot. The response was almost similar. Because, I don’t use either of them very frequently.

The response was like 'a student who has not prepared anything for the exams is scribbling something on the answer papers for the sake of just filling it'. Have you seen such scenarios...?? Here is the one below:

?Your software knows only as much as you use them…

When the input (prompt) is good and precise, you are likely to get more precise output. It is all about prompting them properly and training them which get you the better ROI.

On the other hand, I was trying the same with ChatGPT of OpenAI and not surprisingly, it is cutting down the information in the exact format, sizes and pieces of what I am expecting…Its because it knows me now at least for an year….It was even able to cleverly answer my probing questions on 'How do you know me?'...(2nd screenshot below)

Here is the output from OpenAI ChatGPT:

It has lived with me and spent time with me to understand me better and it vice versa for me too (I know 'How to prompt better in Open AI by now?).

Time teaches | Practice makes more perfect

Having said this, the new friend 'Deepseek' is not going to remain in the same state. It will gain its own market share because of the financial freedom it provides and by its technical ability: leaner models that require far fewer GPUs and drastically lower energy consumption, Deepseek has shattered the assumption that AI’s progress must come with skyrocketing energy demand and deep pockets.

Unlike proprietary models, DeepSeek is open-source, fully licensed and openly discloses its reasoning steps.DeepSeek transparency is crucial in an era when trust in AI is paramount, and it sets a new standard for how AI systems should operate moving forward.

DeepSeek R1 has done what Silicon Valley’s biggest players said was impossible — it made AI lean, efficient and affordable.

DeepSeek R1 is rewriting the rules of the game. Let’s look at the numbers:

• OpenAI’s models cost over $100 per million tokens.
• DeepSeek’s cost? Less than $4 per million tokens.

Above all, what I like the most is the philosophy of Deepseek:

Sometimes, the path to innovation is not about having resources. - It's about using what you have more wisely.

Stay wise, Stay safe & Stay protected

?Whichever be the AI tool that you are using, as user we need to ensure that we follow certain best practices while using them.

Protecting personal identities and data while using AI tools is critical, especially as AI technologies become more integrated into daily life. Here are some best practices to help ensure privacy and security:

1. Use Secure, Reputable AI Providers

• Choose AI tools and service providers who prioritize data security and have a proven track record of complying with data protection regulations like GDPR, CCPA, or HIPAA.
• Review the privacy policies and terms of service of the AI tool providers to ensure they align with your privacy standards.

2. Data Anonymization

• Use anonymization or pseudonymization techniques when processing data. This can significantly reduce the risk of personal data being exposed in case of a breach.
• Where possible, avoid using identifiable information such as names, addresses, or phone numbers in training datasets.

3. Monitor and Manage AI Decisions

• Regularly monitor the decisions made by AI algorithms to ensure there is no unintended bias or misuse of personal data.
• Establish human oversight for AI systems, especially in decision-making processes that may affect people's lives or privacy.

4. Limit Data Collection

• Minimize the amount of personal data shared with AI tools. Only provide the essential information necessary for the tool to function.
• Review permissions regularly and disable any unnecessary data collection or sharing options.

5. Opt for Privacy by Design

• Ensure that privacy is integrated into the AI system’s design and functionality from the outset, not just as an afterthought.
• Privacy by Design principles ensure that AI tools are built with data security and user privacy as fundamental components.

6. Educate Employees and Users

• Train users and employees on best security practices when interacting with AI tools, especially when handling sensitive information.
• Ensure they understand the risks of phishing, social engineering, and other attack vectors that could compromise personal data

7. Implement Strong Authentication

• Use multi-factor authentication (MFA) to protect access to AI tools and the data they store or process. This adds an extra layer of security against unauthorized access.
• Ensure AI tools support strong, unique passwords and MFA for all user accounts.

8. Regular Audits and Monitoring

• Conduct security audits and risk assessments regularly to ensure AI tools are operating within security and privacy guidelines.
• Continuously monitor the AI system for any signs of anomalous behavior or security breaches.

9. Stay Updated on Regulations

• Stay up-to-date with evolving data privacy regulations (like GDPR, CCPA, etc.) to ensure compliance with data protection laws.
• If using AI in different regions, be aware of local privacy laws that may require specific protections or restrictions.

10. Understand AI’s Data Retention Policies

• Delete personal data regularly: Many AI systems store personal data for different purposes, such as learning and improving the AI. Make sure that the tool offers an option to delete your data. Periodically review and delete any information you no longer need or that’s unnecessary for the AI's function.
• Review data retention timeframes: Ensure that the tool's data retention policy aligns with your preferences. Some tools retain data indefinitely, while others may have expiration dates for stored information.

By adopting these practices, you can significantly reduce the risks to personal identities and data while still benefiting from the power and potential of AI technologies.

Always remember: privacy and security are ongoing processes, not one-time fixes.