Twitter, ChatGPT and a Tech Billionaire

What is the actual significance of ChatGPT and its practical applications, ponders Timo Seppälä, lead researcher at ETLA, in his column.

As the Internet becomes increasingly enveloped in the complexities of social media, where does non-social media belong? Should we relocate all non-social media somewhere else such as a new company-specific Internet? Will such a move drive the need for a future ‘TCP/IP protocol’ for industrial communication and content?

In the context of the use of social media, in particular Elon Musk’s Twitter, we have learned to be mindful of automatically generated, incomplete, and frequently misleading information such as fake news. As far as Twitter is concerned, there has also been deep discussion about counterfeit accounts.

Over the past month, the same debate about online veracity has broadened to cover the system of large language models. With the proliferation of systems such as Galactica and ChatGPT, do we now have a new tool for disseminating contemporary content and influencing different actors on the Internet, especially in social media? Along with ChatGPT and Galactica, the Copilot system, previously released for coders, has also astonished with its capability.

Large language models and the content created by them are expected to grow on the Internet.

Of the recent applications for large language models, ChatGPT has captured global attention in the last week. ChatGPT charms with its excellence. Certain testers are convinced of the capability of ChatGPT, while others are less convinced. Still others criticize the superficial results. For now, however, ChatGPT is a research experiment and a showcase of an innovative technology.

As the systemic use of large language models becomes more common, it is important to ask whether their implementation and use will bring wider benefits or disadvantages. We already see the generation of benefits in narrow use cases, however for broader use cases, the situation remains a question mark. There is not enough research data on this yet. For coders, the use of Copilot has been of significant benefit, although there are also still a multitude of question marks surrounding its use, for example regarding copyright disputes.

As a technology pragmatist, may I ask, is this cake rich with substance, or just “beautiful on top, very smooth around the crust, but is sheer inside, chaff underneath the crust?” Will the impending publication of unfinished research results crack the crust and be a big blow to technology companies?

The most important thing is critical reading of search results and source criticism

ChatGPT accelerates the acquisition and assimilation of information in the same way as do the web browsers and search engines of the last thirty years. The difference between the latest technology and its predecessors so far is that the newest large language models does not always reveal the origin of the information, while in the case of older search engines and browsers, sources are known or discoverable. One plagiarizes and the other does not.

So, is the information found in systems such as ChatGPT correct? By extension, if the data driving systems like ChatGPT is collected from Internet, how biased is it? And what of the ethics of using ghostwriters like ChatGPT to create writing? For the time being, the development of large language models should therefore proceed with small steps and healthy pragmatism.

The Internet is becoming a “dump” of information – something should be done!

If we allow systems such as ChatGPT to publish openly on the Internet, we should pay special attention to what is accurate and inaccurate, and more urgently what is right and what is wrong. Is the entire internet becoming a similar data dump like social media has been for a long time? One tech billionaire – Elon Musk – might be on to something. It may be desirable to move towards the wider use of stronger authentication and paid systems, thereby restoring the reliability of information on the Internet to a level where it is worth using it for something other than its original purpose. Keep in mind that the Internet began as just a dumb data transfer pipeline.

In a broader sense, societies must be prepared for new types of attempts to influence companies conducted through misinformation. How is society responding to this challenge? And at last is the power of information increasingly transferred to Digiants?

 

This column was not written by ChatGPT but by an identified person.