Generative AI: the good, the bad, and the ugly

Hello YODLers,

I've just read "The threat from large language model text generators", a technical advisory article by the Canadian Centre for Cyber Security, on the threats posed by generative AI tools based on Large Language Models (LLM).

I leave you with my top three lessons, and with my curiosity to learn how YODLers use LLMs driven apps in your businesses and work, which tools you use, and what are your thoughts about the threats explained in the article.

  1. The Good: Rapid evolution of LLMs has opened the door to versatile applications in content generation and dialogue systems, such as analysis of legal documents, health (including mental health) diagnosis, food recipe development, food trend analysis, nutritional information analysis and generation, and customer service and engagement.
  2. The Bad: AI-generated content is being used by malicious actors to generate and spread misinformation, and influence public opinion through social media, and to launch credible phishing email campaigns.
  3. The Ugly: Preventing the abuse of LLM-generated content is challenging, making it difficult to combat the spread of disinformation, and the potential for misuse in phishing campaigns.
2 Likes

AI seems to be such a buzzword these days. I see people talking about AI features on their apps and websites and often they seem to be just regular old search functions. I think this is another downside to this trend: people trying to capitalize on this trend who are not really adopters. Unfortunately, they seem to be causing confusion for people trying to understand what AI is and how it affects their own lives.

1 Like

AI has its' pros and cons for sure. We have been using AI for over a year now for accelerating the creative process for both design and copy writing saving clients thousands on their packaging design. We have found that it still takes time editing copy after the fact, and many rounds of prompt refinement for image creation. one area that is being adopted are AI Chatbots. These need training for several months before deploying to the public. I was on a site today that had an AI chatbot that gave incorrect answers to all of my questions. AI Chatbots still require training on your product line, FAQs, Services, catalogues, etc etc, before they should be deployed to the public. Sometimes rushing to be a bleeding edge adopter has its challenges. AI is a tool, a toolbox of tools, but they must be used correctly. And there needs to be oversight like a regular employee. For us we have had great results for draft copy - getting beyond the 'white page' that many writers wrestle with. We have also been successful for spot illustrations and 'stock' photos. A recent packaging project for a client for 3 different product lines resulted in several thousands dollars saved, removing several custom illustrations and custom photos being required. Use of AI also cut copy writing costs in half over previous projects. There is a benefit to be had for sure, but there is a learning curve.

1 Like

Hi Nestor,

Good question!

Yes, artificial intelligence (AI) can do many things, but NOT everything. Do you know, with our intellectual property (IP), we can do what AI can NOT do? Here is an experiment on ChatGPT with China's data. This feature can make your information service UNIQUE in the world.

AI is on 2024 Davos World Economic Forum's agenda.

https://lnkd.in/gSM-tXxB

But when we used the following Chinese-English multilingual question to test ChatGPT, an icon of AI, the experiment result showed that it can NOT answer this kind of questions, while we can do, with our IP.

"Who, in the '湖北' province of mainland China, has new US patents granted on January 16, 2024?"

Our IP is a copyrighted multilingual metadata. It can provide the real time (2024.01.16) information about US patent holding/holder, based on the multilingual census geographical locations of Canada, mainland China, Hong Kong, Macao, Taiwan, Middle East (Israel, Saudi Arabia, UAE) and Europe (Austria, Germany, Switzerland, UK).

This proves the criticality and UNIQUENESS of our IP:

Without metadata, especially multilingual metadata, NO multilingual data can be found/retrieved for intelligence analysis, Internet of Things (IoT), etc., even by the most advanced technologies, like artificial intelligence (AI), supercomputers, etc.

https://lnkd.in/g-aJFnXR

I am seeking a copyright licensing buyer for this IP of mine. Please contact me at henry.chang212@gmail.com if our IP rings a bell to you. Thanks.

Regards

Henry

1 Like

Hey @Robert Dodds,

Thanks for sharing your experience with generative AI. I see a lot of similarities with my own experience. I have been using ChatGPT for almost year now with different levels of success, to research technology solutions, troubleshoot and solve technical issues, structuring project proposals, writing summaries, extracting and consolidating ideas, writing speeches, and for analyzing complex documents. In absolutely ALL cases, the product is never final. I always do my due diligence, verify facts, test sequence of instructions, personalize language, and overall, never trust it blindly.

The learning curve hasn't been too difficult and I am enjoying the journey.

My experience generating graphics with prompts has been less satisfying though.

@Nestor Gomez

AI is becoming another high-tech bubble.

IP is an asset that consolidates innovation for marketing and making money, while food technology is a major category of IP (patent). Marketing needs intelligence about strength, weakness, opportunity threat (SWOT).

Our IP is a copyrighted Chinese-English multilingual metadata, sorted by the census geographical locations of US patent assignee in Canada, mainland China, Hong Kong, Macao, Taiwan, Middle East (Israel, Saudi Arabia, UAE) and Europe (Austria, Germany, Switzerland, UK). It is just like a “High-tech IP A-Z of Canada”, “High-tech IP A-Z Page of China”, etc.

It is UNIQUE, and is an enabler and index in applying modern technology, especially artificial intelligence (AI), to do cross-border multilingual open source data search/analytics (OSINT) for SWOT, as in Figure 1.

In short, we are using our own intellectual property (IP), on a low-cost ordinary laptop, to do what high-tech solutions, e.g. AI and quantum computers, can NOT do in multilingual cross-border SWOT data search/analytics.

At this moment, there is NO data solution in the world, artificial intelligence (AI) or NOT, that can answer the following Chinese-English multilingual question of SWOT market intelligence:

“How many entities are there in the Ontario province of Canada, or in the 江蘇 province of China, have new US patents granted on the nearest Tuesday (Eastern Time Zone), when USPTO released the newly approved US patents on a weekly basis?”

With our IP, we can answer such questions, even on a low-cost ordinary laptop without AI.

Please advise your interest and needs by henry.chang212@gmail.com. Thanks.