{"id":240,"date":"2023-07-04T16:51:12","date_gmt":"2023-07-04T14:51:12","guid":{"rendered":"https:\/\/www.mutareb.com\/?p=240"},"modified":"2023-07-05T11:26:35","modified_gmt":"2023-07-05T09:26:35","slug":"chatgpt-for-market-research","status":"publish","type":"post","link":"https:\/\/www.mutareb.com\/index.php\/2023\/07\/04\/chatgpt-for-market-research\/","title":{"rendered":"ChatGPT for Market Research"},"content":{"rendered":"\n

This is a short write up and essentially a summary of prompt engineering techniques, best practices and approaches for market research I came across.<\/p>\n\n\n\n

ChatGPT has also been trained on research papers can can compose well sounding paragraphs. But it should be kept in mind, that it will always try to please the user, which occasionally results in fabricated facts or nonsensical sentences. <\/p>\n\n\n\n

With that said, let’s try to get an initial draft for a market research on AI\/ML in the German automotive sector from ChatGPT 3.5.<\/p>\n\n\n\n

what's the market size of AI\/ML in the automotive sector?<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Well, that’s a bummer, let’s see what data it have anyway:<\/p>\n\n\n\n
what does the latest data you have project?<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Still, nothing useful. This can happen, so I’ll just take another approach:<\/p>\n\n\n\n
who are the main players in Germany?<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
place them in a table with name and focus areas<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
please add their respective revenue to the table<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Not exactly what I expected, but let’s have another try at it:<\/p>\n\n\n\n
add the most recent data you have, with the year in brackets<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Well, sometimes you’ll notice ChatGPT is like a new intern. You still have to rephrase your question and be more specific to get the results you want:<\/p>\n\n\n\n
just give me the data from 2020<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Finally!
But are these numbers accurate? Are they even real? My advice, never take any facts from a chatbot, get them yourself from a reliable source.<\/p>\n\n\n\n
Anyway, let’s sort the table first<\/p>\n\n\n\n
please sort the table by revenue<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Now let’s be more specific, give it a clear task and a context including the expected output<\/p>\n\n\n\n
I want to create a report on AI\/ML in the automotive sector in Germany. \nI am interested in identifying the high growth segments, as well as latest innovations within this space. <\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
This was a good response. Go ahead and read it, sounds very convincing.<\/p>\n\n\n\n
Now let’s ask it to structure the report next:<\/p>\n\n\n\n
I want to focus on analytics, especially customer behavior. \nWhat would you suggest this report should focus on?<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
You can really notice, that it consumed quite some research papers and market studies, it just generated it within a few seconds what would have taken at least 30 minutes of putting the text together alone.<\/p>\n\n\n\n
Now we drill down into more specifics, this approach seems to reduce incoherent sentences and AI hallucinations, as we keep guiding the AI step by step and give it clear instructions in a manageable scope.<\/p>\n\n\n\n
So we ask it now to expand on the first point, make it clear we want 4 paragraphs, to make sure it’s not too short, but again, proof read the output.<\/p>\n\n\n\n
write the first section of this report. make it 4 paragraphs long<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
\"\"<\/figure>\n\n\n\n
It’s amazing how convincing it sounds. All it does is just guess what the next word in a chain of words will most likely by. But it sounds good.<\/p>\n\n\n\n
\"\"<\/figure>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Great. Let’s make this a little bit more professional with a table summarizing profitability and growth estimations<\/p>\n\n\n\n
create a table containing each of these application areas, and provide\n an assessment of profitability and growth within each of the areas<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Looks great, and took only a few minutes.
<\/p>\n\n\n\n
Let’s ask it to rewrite a section based on the last table:<\/p>\n\n\n\n
Please, rewrite the 2nd section, focus only on the areas with both \nhigh profitability and growth assessment<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Go back to the assessment table, only customer experience optimization met the requirements, as mentioned before, as much as it might impress you with how well written the responses are, you have to double check every output, you can always trust it make things up.<\/p>\n\n\n\n
Let’s wrap our research up, we can’t look serious without a list of resources:<\/p>\n\n\n\n
provide a list of possible references for the information contained \nin the previous response. Include a link to the website.<\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
The thing is, none of these links are real. The websites are real, but the papers are about a completely different topic. For instance, the first link takes me to a paper about macrophage migration, also the authors are different<\/p>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Well, at least Vishal Gupta is a real associate professor of Data Science, might be a starting point for finding a more credible sources.
In fact, all of the resources are made up by ChatGPT, some of them even had fictive authors.<\/p>\n\n\n\n
So be aware of this whenever you use a chatbot to generate content!<\/p>\n\n\n\n
Here you have it, a professional looking, well written albeit made up market research of AI\/ML sector in the German automotive industry.<\/p>\n","protected":false},"excerpt":{"rendered":"
This is a short write up and essentially a summary of prompt engineering techniques, best practices and approaches for market research I came across. ChatGPT has also been trained on research papers can can compose well sounding paragraphs. But it should be kept in mind, that it will always try to please the user, which… Continue reading ChatGPT for Market Research<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4],"tags":[49,50,51],"series":[52],"class_list":["post-240","post","type-post","status-publish","format-standard","hentry","category-machine-learning","tag-ai","tag-chatbot","tag-chatgpt","series-prompt-engineering","entry"],"_links":{"self":[{"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/posts\/240","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/comments?post=240"}],"version-history":[{"count":2,"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/posts\/240\/revisions"}],"predecessor-version":[{"id":260,"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/posts\/240\/revisions\/260"}],"wp:attachment":[{"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/media?parent=240"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/categories?post=240"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/tags?post=240"},{"taxonomy":"series","embeddable":true,"href":"https:\/\/www.mutareb.com\/index.php\/wp-json\/wp\/v2\/series?post=240"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}