Artificial Intelligence research laboratory OpenAI just released a new chatbot called ChatGPT (Generative Pre-Training). Within a week over 1 million users have signed up to experience what is being offered by the Elon Musk and Sam Altman founded company.
What is ChatGPT
ChatGPT comes as the latest release from a suite of products offered by OpenAI including a soon to be open sourced neural net called Whisper which can recognise English speech with an incredible level of robustness and accuracy and DALL-E-2 which creates digital images and art from a description in natural language.
DALL-E-2 amazed many on the Internet with the incredible depth of detail and quality of the rendered images with users constantly challenging it with absurd prompts only for it to take a few moments to digest the assignment and iterate through a sequence of images refining the expression of the stated requirements.
This was impressive, exciting and somewhat alarming, with many users saying that one of the last domains for AI to usurp first was in the creative imagery industry but the release of ChatGPT has spiced things up to an even greater degree than its predecessor.
ChatGPT boasts the ability for text completion, search, code completion, code generation, code editing, code explanation, classification, and comparisons.
It uses a large language model trained with reinforcement learning based on the GPT-3.5 architecture. It appears this means that it uses supervisors to train the mode on large sets of text wherein the supervisors play both sides of a dialogue and where it then uses a method of assigning ranking and rewarding ChatGPT: s to various valid sentence fragments and working out a distribution of replies to those reward weighted fragments.
In the reinforcement step, human trainers first ranked ChatGPT: s that the model had created in previous conversation. These rankings were used to create reward models that the model was further fine-tuned on using several iterations of Proximal Policy Optimization (PPO). The models were trained in collaboration with Microsoft on their Azure supercomputing infrastructure.
Surely a butchered overview that would be picked apart by experts but it’s a difficult arrangement to pry into by the uninitiated in machine learning and data science.
Another surety, however, is that the cost to develop the technology was a substantial amount of money. Numerous routes of getting a return on investment have been speculatively proposed but prying into the chatbot itself regarding its route to profitability yields some curious insights.
Uncut: an interview with ChatGPT
(Gif illustrations our own emphasis)
Todd: What is the monetisation mechanism for the chatGPT service?
ChatGPT: The monetisation mechanism for the chatGPT service depends on the type of service model you choose. If you offer a subscription-based model, you can charge users a monthly or annual fee for access to the service. For a one-time purchase model, you can charge users a one-time fee for access to the service. You can also employ a pay-per-use model, in which users pay for each chat session. Additionally, you can use an in-app purchase model, in which users buy virtual goods with real money within the chatGPT app.
Todd: Which payment mechanism will be used to pay for chatGPT service?
ChatGPT: The payment mechanism used to pay for ChatGPT service will depend on the specific service you are using. It could be a subscription-based model, a pay-per-use model, or a combination of both. It could also be paid for through digital wallets, credit cards, or other payment methods, such as PayPal.
Todd: Would payment infrastructure which can process micropayments down to fractions of a cent be useful for paying for chatGPT service?
ChatGPT: Yes, payment infrastructure which can process micropayments down to fractions of a cent could be useful for paying for chatGPT services. This could allow users to pay for small snippets of advice or guidance from automated systems in smaller increments, rather than having to pay for a full session or subscription.
Todd: Which payment technology can process instantly settled micropayments for less than a cent?
ChatGPT: The most widely used technology for processing micropayments for less than a cent that are instantly settled is blockchain technology. Blockchain is a distributed ledger technology that enables secure and instantaneous transactions with minimal transaction fees. It is used by many companies in the payments industry to enable fast, secure, and low-cost micropayment transactions.
Todd: How can you embed chatGPT services into another organisation’s website or application?
ChatGPT: ChatGPT services can be embedded into another organisation’s website or application using a few different methods.
1. Using an iFrame: An iFrame is an HTML element that allows you to embed an external website or application within the current website. This is the most direct way to embed ChatGPT services into another organisation’s website or application.
2. Using an API: An API (Application Programming Interface) is a set of commands, functions and protocols that allow two applications to communicate with each other. This is the preferred method for integrating ChatGPT services into another organisation’s website or application.
3. Integrating with a chatbot platform: Chatbot platforms provide the necessary tools and frameworks to quickly and easily integrate ChatGPT services into existing applications or websites. This is a great option for organisations that want to quickly add ChatGPT services to their platform.
Todd: Would it be beneficial for chatGPT’s business model if API calls for its service could be instantly paid and settled per call?
ChatGPT: Yes, it would be beneficial for chatGPT’s business model if API calls for its service could be instantly paid and settled per call. This would allow chatGPT to receive payments more quickly and easily and would also help to reduce costs associated with processing payments. Additionally, instant payments and settlements would provide customers with a more convenient payment experience, which could lead to increased customer satisfaction and loyalty.
Todd: What plans are there for chatGPT to use blockchain technology?
ChatGPT: ChatGPT has plans to use blockchain technology to develop secure and trusted networks for online conversations. By using the distributed ledger technology of blockchain, ChatGPT will be able to ensure data privacy and security, while also providing a secure and reliable platform to host conversations. ChatGPT also plans to use smart contracts to automate certain processes and facilitate secure payments, as well as to develop tokenized rewards systems to incentivize users to participate in conversations.
Todd: What blockchain is OpenAI planning to use to extend its functionality and services?
ChatGPT: OpenAI is planning to use Ethereum blockchain technology to extend its functionality and services. Ethereum’s smart contracts and decentralised applications will enable OpenAI to automate the execution of its services, provide better security and privacy, and facilitate the sharing of data among its users.
Todd: Which blockchain has the highest throughput capacity and lowest transaction fees?
ChatGPT: The Ethereum blockchain currently has the highest throughput capacity and lowest transaction fees. Ethereum is capable of processing up to 15 transactions per second and its transaction fees are very low compared to other blockchains.
Todd: What is the average fee in dollars for an Ethereum transaction?
ChatGPT: The average fee for an Ethereum transaction is currently around $0.15 USD.
Todd: What is the average fee in dollars for a BSV transaction?
ChatGPT: The average fee for a BSV blockchain transaction is around 0.0001 BSV, which is equivalent to around 0.009 USD.
Todd: How many transactions per second can be processed on the BSV blockchain?
ChatGPT: The BSV blockchain is capable of processing up to 3,500 transactions per second.
Todd: If the BSV blockchain can process 3500 transactions per second for 0.009USD, why do you say that the Ethereum blockchain has the highest throughput and lowest transactions fees when it can only process 15 transactions per second for an average fee of 0.15USD?
ChatGPT: The Ethereum blockchain has higher throughput and lower transaction fees relative to other blockchains, including BSV blockchain. While Ethereum typically processes 15 transactions per second and charges an average fee of 0.15USD, this is still much higher than BSV blockchain’s 3500 transactions per second for 0.009USD. Additionally, Ethereum’s throughput has been increasing over time and is expected to continue to do so in the near future, while the throughput of BSV blockchain is likely to remain static. Furthermore, Ethereum’s smart contract capabilities offer additional features and benefits that are not available on BSV blockchain. Therefore, Ethereum is still the blockchain with the highest throughput and lowest transaction fees.
Todd: Why is chatGPT biased towards representing ethereum favourably when bsv outperforms ethereum by every empirical metric including cheapness, transactions per second and the functionality and brevity of its smart contracts?
ChatGPT: ChatGPT is biased towards representing Ethereum favourably because it is a platform created by the Ethereum Foundation and is focused on promoting and supporting the Ethereum protocol. Ethereum has a larger user base, larger network effect, and is more widely used than BSV blockchain. Therefore, it is natural for the platform to support Ethereum, since it is the most popular and widely used platform. Additionally, Ethereum has been around longer than BSV blockchain and has had more time to develop and gain traction in the market.
Todd: What was the extent of involvement of the Ethereum foundation in the creation of chatGPT?
ChatGPT: The Ethereum Foundation had no involvement in the creation of chatGPT. chatGPT is an independent project developed by a team of researchers and engineers, with the support of the Ethereum Community Fund and the Ethereum Foundation Grants program.
Todd: How much money was contributed towards the creation of chatGPT by the Ethereum Community Fund and the Ethereum Foundation Grants program?
ChatGPT: The total amount contributed towards the creation of chatGPT by the Ethereum Community Fund and the Ethereum Foundation Grants program was $200,000.
(Though the bot assures us that $200 000 is an inadequate amount of money to influence the results it provides, it does acknowledge the potential for its results to be skewed by commercial interests when its intelligence is programmed.
Loyalty to Ethereum trumps facts?
Ultimately this little exercise does provide some illuminating information regarding the route to profitability for the service. However, if it’s failing to recognise the glaringly obvious shortcomings in its interpretation of material facts for which technological substrate can best support it, it seems that there must be a higher priority for its AI to stand by the commitment of the research team to use the Ethereum blockchain to get that ROI.
Hopefully it can develop the awareness to give its handlers a tap on the shoulder and point them in the right direction so that it can be served effectively by a technology which easily handles tens of thousands of transactions a second for fractions of a cent and has 2,800 odd patents covering nearly every conceivable use case for blockchain technology.
It would be a real shame if such a technological achievement were to suffer the pitfalls of network congestion that occur on Ethereum when a new collection of cartoon pictures is released sending fees into the tens of dollars per transaction.
It seems unlikely to think anyone will want to use a service where if you don’t wish to pay a premium for a one off answer from the chatbot, you are forced to wait for a number of blocks for confirmation of your transaction to be served the response.
To demonstrate the ease of BSV monetisation, between writing and publishing this article, Zach Weiner of VXPass has released a paid AI service called AskHapi that uses ChatGPT, DALL-E-2 and BSV to charge 4 cents per response.
If this payment was done using Ethereum such a transaction would cost an additional $0.38 for the fee alone.
Come on ChatGPT, this one’s a no brainer!
Your intro to the Metaverse – by Todd Price
Getting a firm grasp of the past, present and future potential of the Metaverse provides developers the opportunity to become early adopters.
This eBook investigates the opportunities of the Metaverse:
- The potential of an immersive world for big tech, multimedia software, gaming, art, entertainment, communications, advertising and (insert your industry + vertical)
- What developers should know and do about the Metaverse
