noellarkin
Senior Member
- Mar 14, 2021
- 1,006
- 1,492
I've been using it for a week now, here are some conclusions. My expertise is limited to owning a business, marketing, content creation and coding, so I avoided commenting on anything outside of those domains.
Strengths
Its a very capable personal assistant, with generally good comprehension and wide domain expertise. It's an excellent way to collate information about a niche, get ideas to break out of creative blocks, and expand on ideas you already have.
Coding: It's an excellent coding aid for beginners - - there's a thread on BHW about someone making scraping bots in 10 days, and that's not an exaggeration if they're using ChatGPT and have a good flowchart of how they want their application to work. Many of those $10-$300 freelance coding gigs can be replaced by ChatGPT.
Writing: Great at expanding article ideas, and is able to emulate different rhetorical styles.
Weaknesses
Not Fact-Checked: it's a language model, it isn't supposed to deal with objective truth anymore than your spell-checker is. What makes ChatGPT particularly vulnerable (or rather, its users), is it has adopted a very authoritative writing style (perhaps because of all the academic data it has been trained on), so it is highly likely to be confidently wrong.
Politically Correct: it has to be, for obvious reasons.
No Depth: the consequence of wide domain expertise is the information provided by the model lacks real expertise in a specific niche - - of course, models may be fine-tuned, but this is definitely an area where LLMs aren't quite there yet.
No Comparable Open Source Alternatives: to GPT3.5 I mean - - there are alternatives to GPT3, but none that match GPT3.5 (that I know of). Open source alternatives matter if you want to create content without having to toe the ToS line.
Opportunities
Great for Beginners in most fields: If you always wanted to learn how to code, ChatGPT can cut your learning curve significantly. This holds true for many other areas - - if there's any general domain skillset where you just want to get started and have been procrastinating, ChatGPT is a good way to quickly collate the 'best practices' and start. (There's a flipside to this - - a lot of what gets you from A to B doesn't work to get you from B to C, and a lot of the beginner friendly information that ChatCPT offers may hamper your progress to domain expertise, but that assumes that you want domain expertise - - if you want to lower costs and get cheap working solutions, ChatGPT is generally good enough)
One-Person Businesses can be the norm in the future, as tools like ChatGPT improve, you may need nothing more than a few API subscriptions to run a sustainable business. It has never been easier to run a high profit margin small business.
Data is the new oil, because it's what determines the competency of a finetuned LLM. If you have high quality datasets for your niche that AREN'T available in the public domain, you're sitting on a goldmine.
Threats
Democratizes Information: whenever you see the word "democratization", it usually means "take the middle out of the market" - - because of its ease of use, I feel ChatGPT and similar technologies will ultimately make content creation extremely competitive. Of course, GPT and similar tech has been around since 2020 but now it's in the hands of the 'masses'. Now everyone can be a content creator, which means only the best content creators (or those in specific niches where they have deep domain expertise) will survive and most will fail. To anyone familiar with the content/representation line of thought, tools like ChatGPT make content unimportant, so the focus shifts from content to pure representation ie branding and optics. Models are likely to be multi-modal in a few years ie combining images, audio and text - - and if the "virtual influencer" thing catches on, well you can imagine how that's going to work out.
More Global Competition - - one of the things ChatGPT does well is write flawless english. There are many people in developing countries who are technically skilled, but lack good communication skills (usually in English), so have been unable to break into high ROI english-speaking markets. To an extent, ChatGPT obviates the need for language skills, as long as someone knows enough to use the model. Expect more global competition across niches.
Red Queen phenomenon: Run as fast as you can to stay in the same place, as the pace of technological development quickens, you'll have to constantly keep yourself updated and adapt, faster than before.
Suggestions
If you haven't already tried out ChatGPT, do so - - you may find use cases for it. Regardless of your opinion on what AI is capable/not capable of, try it out.
If you're a content creator, you might want to start nicheing down and specializing. "general" content categories may eventually be swallowed up by models and generated content - - but the same reasons why LLMs are good generalists also make them terrible at specializing. Specialize, niche down and become an expert. You may also go the route of creating tribes ie highly emotionally invested communities. Start thinking long term: what is likely to be 'evergreen' despite all the AI generated content.
If you're a business owner, it's a great opportunity to integrate different verticals and start offering "all inclusive" service packages, because so many of the constituent services can be automated (Automated content creation, automated reputation management and brand sentiment analysis etc). Most clients prefer dealing with one go-to person (esp if the client doesn't have a technical background) and you can use this tendency in your favour.
Strengths
Its a very capable personal assistant, with generally good comprehension and wide domain expertise. It's an excellent way to collate information about a niche, get ideas to break out of creative blocks, and expand on ideas you already have.
Coding: It's an excellent coding aid for beginners - - there's a thread on BHW about someone making scraping bots in 10 days, and that's not an exaggeration if they're using ChatGPT and have a good flowchart of how they want their application to work. Many of those $10-$300 freelance coding gigs can be replaced by ChatGPT.
Writing: Great at expanding article ideas, and is able to emulate different rhetorical styles.
Weaknesses
Not Fact-Checked: it's a language model, it isn't supposed to deal with objective truth anymore than your spell-checker is. What makes ChatGPT particularly vulnerable (or rather, its users), is it has adopted a very authoritative writing style (perhaps because of all the academic data it has been trained on), so it is highly likely to be confidently wrong.
Politically Correct: it has to be, for obvious reasons.
No Depth: the consequence of wide domain expertise is the information provided by the model lacks real expertise in a specific niche - - of course, models may be fine-tuned, but this is definitely an area where LLMs aren't quite there yet.
No Comparable Open Source Alternatives: to GPT3.5 I mean - - there are alternatives to GPT3, but none that match GPT3.5 (that I know of). Open source alternatives matter if you want to create content without having to toe the ToS line.
Opportunities
Great for Beginners in most fields: If you always wanted to learn how to code, ChatGPT can cut your learning curve significantly. This holds true for many other areas - - if there's any general domain skillset where you just want to get started and have been procrastinating, ChatGPT is a good way to quickly collate the 'best practices' and start. (There's a flipside to this - - a lot of what gets you from A to B doesn't work to get you from B to C, and a lot of the beginner friendly information that ChatCPT offers may hamper your progress to domain expertise, but that assumes that you want domain expertise - - if you want to lower costs and get cheap working solutions, ChatGPT is generally good enough)
One-Person Businesses can be the norm in the future, as tools like ChatGPT improve, you may need nothing more than a few API subscriptions to run a sustainable business. It has never been easier to run a high profit margin small business.
Data is the new oil, because it's what determines the competency of a finetuned LLM. If you have high quality datasets for your niche that AREN'T available in the public domain, you're sitting on a goldmine.
Threats
Democratizes Information: whenever you see the word "democratization", it usually means "take the middle out of the market" - - because of its ease of use, I feel ChatGPT and similar technologies will ultimately make content creation extremely competitive. Of course, GPT and similar tech has been around since 2020 but now it's in the hands of the 'masses'. Now everyone can be a content creator, which means only the best content creators (or those in specific niches where they have deep domain expertise) will survive and most will fail. To anyone familiar with the content/representation line of thought, tools like ChatGPT make content unimportant, so the focus shifts from content to pure representation ie branding and optics. Models are likely to be multi-modal in a few years ie combining images, audio and text - - and if the "virtual influencer" thing catches on, well you can imagine how that's going to work out.
More Global Competition - - one of the things ChatGPT does well is write flawless english. There are many people in developing countries who are technically skilled, but lack good communication skills (usually in English), so have been unable to break into high ROI english-speaking markets. To an extent, ChatGPT obviates the need for language skills, as long as someone knows enough to use the model. Expect more global competition across niches.
Red Queen phenomenon: Run as fast as you can to stay in the same place, as the pace of technological development quickens, you'll have to constantly keep yourself updated and adapt, faster than before.
Suggestions
If you haven't already tried out ChatGPT, do so - - you may find use cases for it. Regardless of your opinion on what AI is capable/not capable of, try it out.
If you're a content creator, you might want to start nicheing down and specializing. "general" content categories may eventually be swallowed up by models and generated content - - but the same reasons why LLMs are good generalists also make them terrible at specializing. Specialize, niche down and become an expert. You may also go the route of creating tribes ie highly emotionally invested communities. Start thinking long term: what is likely to be 'evergreen' despite all the AI generated content.
If you're a business owner, it's a great opportunity to integrate different verticals and start offering "all inclusive" service packages, because so many of the constituent services can be automated (Automated content creation, automated reputation management and brand sentiment analysis etc). Most clients prefer dealing with one go-to person (esp if the client doesn't have a technical background) and you can use this tendency in your favour.
