Working with GPT3 via Neuroflash and DALL-E2 Vs ChatGPT (Case-studies)
Aktualisiert: 7. Feb.
Since 2007, Mr. Campaigning has been providing communications support to the Swiss pension fund association ASIP. Always on the road in the interest of the insured, Mr. Campaigning relies on the most innovative technology.
Efficient, economical but effective use of resources is an essential aspect in the planning and implementation of communication measures.
ASIP began using social media as early as 2010. They had a chatbot answer questions about AV2020 in 2017 and commissioned AI-supported market analyses from evAI Intelligence GmbH in order to keep its finger on the pulse of communication with its various stakeholder groups.
Copyright-free images using DALL-E2 AI Network
In recent months, the AI networks GPT3 and DALL-E2 have also been increasingly used.
DALL-E2 created its own new images that Mr. Campaigning uses to illustrate blog articles without fear of copyright infringement.
For example, for the article "BVG-Reform: Kritik von allen Seiten" the briefing for the AI was "a tree split by several lightning bolts in comic style",
For the article "Mit blauer Ökologie gegen die grüne Ökodiktatur" it was "blue ecology".
Why is ChatGPT excellent for explaining facts? Case study
On August 31, 2022, Mr. Campaigning wanted to publish ASIP's position based on media reports on ESG, while also explaining what ESG is. Because the matter was urgent and the author had only 15 minutes until the next deadline, he used his account at Neuroflash.
This platform, like the currently hyped ChatGPT, accesses OpenAI's GPT3 network and produces texts for such diverse applications as blog articles, slogans, Google Adwords texts, etc.
Mr. Campaigning gave a briefing on "The meaning of ESG criteria and their meaning". The result was an explanatory text that perfectly complemented the association's statement. Quality control was carried out by the association's director, who gave the green light for publication within a few minutes. By then, 10 minutes had passed.
It was only after the check that the director learned that artificial intelligence had written the explanatory text. He had not noticed it - another proof of quality. If the text had seemed strange without the director knowing where it came from, that would have been a sign of the need for revision.
As it was, however, it went through without a hitch and was published, "Die Bedeutung von ESG-Kriterien und ihr Sinn"
Using Neuroflash for website copy
On the website of Mr. Campaigning, there are several explanatory texts written by Neuroflash, for example, the explanation of reputation management, public affairs, or leadership and change management. The text on the website of Mr. Campaigning had to be improved even more than the text about ESG.
This could be due to the fact that ESG is a clearly defined topic area, about which much is written, but there is also widespread agreement on its importance. The situation is quite different when it comes to reputation management, public affairs, and change management. Here, there is a proliferation of definitions because every provider of services in this area tries to shine with its own definitions. This makes it harder for a neural network to find the most appropriate definition.
Designing speeches with Neuroflash
Mr. Campaigning also uses Neuroflash to design speeches. For example, when he was asked to give a speech at the opening of an insurance branch, Neuroflash provided a suitable text about "the importance of insurance for the economy and society" (briefing). This only had to be "spiced up" with a few personal remarks and ready was a speech that not only earned a lot of applause. No one who was asked could have said which part of the speech had been written by an AI. Not even the client. The main thing was to be satisfied and correct in terms of content. But the latter still has to be verified by humans.
Neuroflash Vs ChatGPT
This is shown by the last example. The question "Who is Peter Metzinger?" Both ChatGPT and Neuroflash answer with a completely wrong text, probably due to confusion with the philosopher Thomas Metzinger. Why Peter Metzinger is supposed to have been born in Tübingen, however, remains a mystery.
Is ChatGPT a serious threat to Google?
Conclusion: contrary to current claims that ChatGPT poses a serious threat to Google, the application is ill-suited for searches that are supposed to deliver correct results. The better application is to have steeples written for texts and speeches.
Learn more about Mr.Campaigning and support his entry to the cantonal council campaign here