Notes from a CTO #6: Docsumo Socials, Nepal as Tech-hub, GPT4 & chatGPT
Lets talk about Docsumo's Socials event where 100+ participants showed up and GPTs.
Hey, it’s Bkrm from Docsumo.
You’re reading the sixth issue of Notes from a CTO: my raw canvas of thoughts and collection of interesting resources I found online. My goal here is to spread knowledge at scale. If this is up your alley, have a read and let me know what you think!
1. Running a startup
This week, let's discuss our Docsumo Social event we hosted on February 25th.
The main purpose of the event was to gather like-minded individuals in one place to discuss and get to know each other. One of the activities was a discussion on "How to Make Nepal a Tech Hub?". We had an awesome time discussing and hearing these ideas. Here is a summary:
Government Policy:
Payment Infrastructure: A robust payment infrastructure is a fundamental requirement for building products for the world. It is very difficult to make and receive cross-border payments, and this is one of the basic pains that every entrepreneur faces. Ensuring a robust payment infrastructure will allow companies and individuals to transact easily and efficiently, which in turn can help drive the SaaS and services ecosystem. A well-functioning payment infrastructure is crucial for any country that wants to become a tech hub, as it enables businesses to scale globally and tap into international markets.
The IT sector as a priority: The government can prioritize the IT sector by investing in it, promoting it as a key industry, and ensuring that it receives the necessary funds to support its growth. This can help the country find sustainable sources of foreign reserves, which in turn can help stabilize the economy and reduce its reliance on remittances. By making the IT sector a priority, the country can position itself as a tech hub and attract foreign investment and talent.
Easiness of doing business: Nepal's ranking of 94 among 190 economies in the ease of doing business is not favorable and needs to be drastically improved. For instance, yesterday, I had to spend three hours at a company office for a task that should have been a simple one-click online job. Simplifying business procedures, reducing red tape, and streamlining government processes can create a more conducive environment for businesses to operate in Nepal. T
Individual Level:
Participation in events and programs: Participation in tech events, hackathons, coding challenges, and other tech-related programs can provide individuals with opportunities to network, learn new skills, and gain valuable experience. This can help Nepal count among the top talent in the world, ensuring that any project developed in Nepal or products built by Nepali talent reflect high-quality standards.
Open source contribution: Contributing to open-source projects can help individuals improve their coding skills, showcase their talent, and build a portfolio of work. This can put Nepal on the map as a talent hub on a global scale, attracting more attention from international companies and investors. By participating in these events and contributing to open-source projects, individuals can hone their skills and develop a competitive advantage that will help them succeed in Nepal's growing tech industry.
College Level:
Updated Course: Colleges can keep their curriculum up-to-date by offering courses in the latest technologies and industry-relevant skills. This will help to produce graduates that are better equipped to enter the job market.
Talent relevant to the industry: Colleges can collaborate with companies and industry experts to develop a curriculum that prepares students for the workforce. By producing graduates with relevant skills, colleges can help create a talent pool that can meet industry demands.
Company Level:
Events and Community: Companies can organize events and create communities to foster collaboration, learning, and growth. This can help to build a positive reputation and attract talent.
Employee collaboration and satisfaction: Encouraging collaboration between employees, and companies and promoting a positive work culture can help to retain talent and improve job satisfaction.
Sharing of knowledge: Companies can create knowledge-sharing programs like blogs, encourage mentorship, and provide opportunities for professional development. It was great to see Khalti's engineering blog. We are starting soon.
Research and Development:
From college, company, and government: Research and development activities can help to drive innovation and solve real-world problems. Collaboration between colleges, companies, and the government can create opportunities for research and development to flourish. This can lead to the creation of new products and services that drive economic growth. We have started by sponsoring a thesis on Pulchowk and KU. Hope to do more in the future.
Capital Market:
VC network: No of VC networks in Nepal can be counted in the finger. Establishing a robust VC network in Nepal can help to provide funding to entrepreneurs and drive innovation.
Angel network: Angel investors are wealthy individuals who invest in early-stage startups. Angel networks can help to support the growth of startups and provide access to capital that may be otherwise hard to come by.
In conclusion, implementing the strategies mentioned above can help create a robust tech ecosystem in Nepal that supports innovation, growth, and prosperity. It is important for all stakeholders, including the government, individuals, colleges, companies, and investors, to work together to establish Nepal as a hub for technology and entrepreneurship. Efforts must be made at all levels, and we cannot just point fingers at each other. We need to take steps at an individual level and work collaboratively to create a thriving tech industry in Nepal. By doing so, we can create new job opportunities, boost economic growth, and help position Nepal as a leading player in the global technology landscape.
Some Pictures from events:
2. Technology
You might be tired of seeing all the posts about ChatGPT and now GPT4, but let me throw my hat in the ring.
Last night, I was thinking about what people felt when they used a computer for the first time. Perhaps we are having a similar experience right now with generative models. Many industries and people's habits will change. In the last two months, I have undergone a lot of changes myself. I went from dismissing the idea that ChatGPT would replace Google in my first newsletter, to using ChatGPT on the command line to search before resorting to Google.
This week at Docsumo, we had a session on LLM models to share the research we have been doing. This model excels in use cases such as solving problems, writing summaries, and finding answers better than 95% of humans. With GPT4, it's in the top 90th percentile on most competitive examples. This model has come a long way, and I think it will be super helpful, much like computers have been. However, will it replace us? It will be more like we have one more superpower in our hands, like a computer. I read somewhere that the PS2 was more powerful than the computer that sent us to the moon, but what are we doing with our computers?
One use case I have been using the hell out of is learning/improving new skill sets, like building a GO ChatGPT cli to help me fix grammar and spelling on email, or solving developer problems that used to take ages for me to search on Google. This is an exciting time.
At Docsumo, we are excited about the improvements in this field and constantly researching how it can help our customers. We have a few ideas in place that are being researched, and a few more that we are just trying to wrap our heads around. If you want to contribute, please apply through the link above.
3. Podcasts
This time, let’s read. Here is the GPT-4 Technical Report.
I understand that, due to the competitive landscape, openAI might not want to share architecture (including model size), hardware, training compute, dataset construction, training method, or any related details. But we expected more detail from open AI. Regardless, it is a must-read for anyone interested in this subject.
If you are wondering what model is used for visual Q&A, this paper(Flamingo: a Visual Language Model for Few-Shot Learning) from Deepmind might give you an idea.
And here's an interesting Twitter conversation. Plot twist: they didn’t disclose any details because there was actually no innovation to report, just a bit more finetuning, which would have looked underwhelming given all the hype.
4. Interesting links
Repos:
Cgpt(Chat GPT on CLI): Over the last week I build cli tool to call chatgpt API and get output on CLI.
Why not use chatGPT UI?
I love cli and chatGPT UI keep going over capacity.
You can set system context on api.
This will be dirt cheap compare for $20 for plus(Now plus give GPT4 access too so might be worth it).
Get better at Go using chatgpt.
llama.cpp: Run llama on cpu. I tried it over this week and the results were just ok. Not a gpt3 replacement for sure but I am really excited about what the future and open source hold. Resource requirement for Linux version ( this might change as this space is evolving very rapidly) :
- 64GB RAM, 8 VCPU, 100GB SSD for 7B
- 128GB RAM, 8 VCPU, 400GB SSD for 65B
neon(Serverless Postgres): I enjoy building hobby projects, and with ChatGPT, I believe I can expedite the process even further. Firebase or superbase for auth and storage, Go or Python for backend, React for frontend, missing: good serverless DB. Have heard good works about Neon. Hope to try soon.
For more follow me on Github: bkrmdahal
Articles:
AI‘s Instagram Problem: Just because someone else has a cool AI project doesn't mean that your project is any less valuable. It's okay to draw inspiration from Instagram, but don't make life-changing decisions just because of FOMO (fear of missing out). Instead, take the time to learn about the technology in detail, including its advantages, limitations, and how to use it effectively.
What Is ChatGPT Doing … and Why Does It Work? For you to help with the first article. Here is one more basic of ML: The rule of machine learning. And million dollar question Will AIs Take All Our Jobs and End Human History—or Not?
The local minima of suckiness: Very Very Very few people are rockstars on any job by birth. The vast majority are self-made through dedication, discipline, and perseverance. To be better developers you need room to make mistakes, repeated exposure to best practices, an understanding of how to ask good questions, and teamwork.
How slack decides to send a notification: Should not this be simple things but check out the workflow? Like Nishan used to say “simple is not easy”.
5. Quotes/ Books
Indeed, the ability to reserve our fucks for only the most fuckworthy of situations would surely make life a hell of a lot easier.
- Mark Manson (The Subtle Art of Not Giving a F*ck)
Life is tough, and overthinking can make it even tougher. As a startup founder, it's common to feel the need to analyze every decision and plan for every possible scenario. However, doing so can create unnecessary mental pressure. Instead, focus on hiring competent heads of department and team leads to share the burden of decision-making. Allow your team to think critically before involving yourself, and trust in their abilities to make informed decisions. Remember, that's why you have a team in the first place!
Our Slack channels have a great collection of memes.
That’s it for this edition, I hope you find it useful.
P.S. I am traveling to the USA this week. Would love to catch up if are near NY(New Jersey) or SF (San José).
Best,
Bikram Dahal
P.S. If you learned something new today, please share “Notes from a CTO” with your friends and spread the love. ✌🏻