What is ChatGPT?

4 min


Information is power, and in this era of technology, the person who holds the keys to information, can guide the narrative.

This the market in which ChatGPT is trying to conquer what, the main competitor to Google search, which has dominated the information marketplace for years without any tangible competitors. That is until ChatGPT came into play and is eating up the search market share in droves.

What is ChatGPT?

ChatGpt is an artificial intelligence language model chatbot assistance, which is able to answer questions queries, from its database of human trained questions and computer generated answers.

ChatGPT is able to sift through these large database of information, to an extent give an accurate answer to the question query asked.

Nevertheless, ChatGPT in its infancy is a chatbot, unlike Google search which uses Ai to give the best website answers to question queries, and it depends on its database which continues to grow by the day.

Also, it incorporates Reinforcement Learning with Human Feedback (RLHF), to improve its database and question queries, and reply answers.

Who Built ChatGPT?

ChatGPT was built by OpenAi, and is a sibling of another language model, instructGPT, these.

OpenAi is a reseach and development company in San Francisco. It was founded in 2015 by Twitter owner Elon Musk, Greg Brockman, Ilya Sutskever, Wojciech Zaremba, and Sam Altman.

Although Elon Musk resigned from the board of directors in 2018, but supports the cause of the company through donations and investments.

What can ChatGPT do?

Essentially ChatGPT can answer most of all questions thrown at it, and even difficult coding questions.

How Was ChatGPT Trained? How does ChatGPT Work (Methods)?

ChatGPT model was trained using by humans and data from web. Especially Reinforcement learning from Human feedback (RLHF), similar to how InstructGPT, but with slight differences in acquiring or collecting data.

Limitations

Like all things built by humans, errors due occur, and in the case of ChatGPT they are limitations of their code:

1. Plausible sounding but error answers

  • Since chatGPT is essentially artificial intelligence, it sometimes writes plausible-sounding answers to the untrained in the field. But on further investigations, it is discovered that these answers were in fact nonsensical.

Resolution Dilema:

  • This error arises from the fact that during RL training, there were no guidelines coded in place, to ascertain, what was true, and what isn’t.
  • Nevertheless, if the artificial intelligence was trained to be cautious, it declined to answer questions, which it could answer correctly.
  • Specially trained individuals were put in place to supervise the ChatGPT ai, but this caused a problem, being that it misled the ai, since the ideal answers depended on what the ai model knows, rather than what the human supervisor demonstrator knows.

2. Sensitivity

  • chatGPT is sensitive to word rephrases due to the vast nature of language models, with idioms, synonyms, and antonyms. As such, the code base was sensitive to little tweaks to the input phrasing, or attempting the same question query phrases multiple times.

Resolution Dilema:

For example, given the ai chat bot a question query, the artificial intelligence model, claims not to know the right answer, but by tweaking that same question query, by rephrasing it, it answers the question query correctly.

3. Excessively verbose

  • In this day and age, time is of the essence, especially when reading. chatGPT has found a way to be excessively verbose, and answers questions with more words, than is needed, instead of being concise and direct to the point.

Resolution Dilema:

These issues arose from biases in training data. Since it is generally accepted that longer answers filled with fluff content, were preferred, and they give the impression of being an intellectual.

4. Ambiguity Numbness

  • ChatGPT was numb to ambiguous queries.

Resolution Dilema:

In other words, instead of asking for more inputs in questions queries it didn’t understand, it just answered the questions anyways.

4. Inappropriate requests Acceptance

  • Despite the fact that efforts has been made for ChatGPT to refuse inappropriate requests. It defied ai orders, and responded to harmful instructions or exhibited biased behavior.

Resolution Dilema:

Currently this behavior is be monitored through the moderation api, to warn or block certain types of unsafe content. Nevertheless, false negatives and positive, will be monitored and the system will be improved.

Is ChatGPT Free to Use?

ChatGPT is currently free to use, but with the pricing page on OpenAi, eventually we think it will become a paid feature.

Eventually, it might become a paid feature.

If we go by the pricing page of the openai page, then the following will be the pricing.

Image models

Build DALL·E directly into your apps to generate and edit novel images and art. Our image models offer three tiers of resolution for flexibility.

RESOLUTIONPRICE
1024×1024$0.020 / image
512×512$0.018 / image
256×256$0.016 / image

Language model

Fine-tuned models

Create your own custom models by fine-tuning our base models with your training data. Once you fine-tune a model, you’ll be billed only for the tokens you use in requests to that model.

MODELTRAININGUSAGE
Ada$0.0004 / 1K tokens$0.0016 / 1K tokens
Babbage$0.0006 / 1K tokens$0.0024 / 1K tokens
Curie$0.0030 / 1K tokens$0.0120 / 1K tokens
Davinci$0.0300 / 1K tokens$0.1200 / 1K tokens

Base models

Ada 

Fastest

$0.0004  / 1K tokens

Babbage 

$0.0005  / 1K tokens

Curie 

$0.0020  / 1K tokens

Davinci 

Most powerful

$0.0200  / 1K tokens

How much did Microsoft invest in OpenAI?

Since its successful launch in 30th November 2022.

Microsoft Corp (MSFT.O) is in talks to invest $10 billion in ChatGPT-owner OpenAI as part of funding that will value the firm at $29 billion, Semafor reported on Monday, citing people familiar with the matter.

Conclusion

Having tested the tool, I am sure this is the tool of the future.

Companies that have failed to adapt and adopt have been left behind in the dust, so in time, we will be seeing more of artificial intelligence (ai), in our work places.

Artificial intelligence shouldn’t be shuned but adopted, as long as adequate security features have been put in place.

After all, we don’t want to create an age of Ultron in our world, since we don’t have avengers to defend us.

Nevertheless, artificial intelligence, can tremendously benefit us, by reducing our workload, and increasing our lifespan by reducing daily stress on out bodies.

See Also

Convert .Crt Certificates

Convert .crt to .Cer

Convert .crt to .jks

Convert .crt to .pem

Convert .crt to .der

Convert .crt to .p7b

Convert .crt to p12

Convert .Cer Certificates

Convert .cer to crt

Convert .cer to .jks

Convert .cer to .pem

Convert .cer to .der

Convert .cer to .p7b

Convert .cer to .p12

Convert .jks Certificates

Convert .jks to cer

Convert .jks to .pem

Convert .jks to .der

Convert .jks to .p7b

Convert .jks to .p12

Convert .jks to .crt

Convert .pem Certificates

Convert .pem to .crt

Convert .pem to .jks

Convert .pem to .cer

Convert .pem to .der

Convert .pem to .p7b

Convert .pem to .p12

Convert .der Certificates

Convert .der to .pem

Convert .der to .crt

Convert .der to .jks

Convert .der to .cer

Convert .der to .p7b

Convert .der to .p12

Convert .p7b Certificates

Convert .p7b to .der

Convert .p7b to .pem

Convert .p7b to .jks

Convert .p7b to .cer

Convert .p7b to .crt

Convert .p7b to .p12

Convert .p12 Certificates

Convert .p12 to .crt

Convert .p12 to .p7b

Convert .p12 to .der

Convert .p12 to .pem

Convert .p12 to .jks

Convert .p12 to .cer


0 Comments

 

7 BUSINESS SECRETS

GET STARTED

This will close in 0 seconds