I've been increasingly worried that AGI is approaching too fast for us to prepare for it. But I thought we still had several years available to prepare. Seeing this, I think we're pretty much there already.
Building a Virtual Machine Inside ChatGPT
411–420 of 947 posts
Re: Building a Virtual Machine Inside ChatGPT
#412Earlier quoted context omitted.
Very cool idea. That'd be a good way to see how it can generalize, actually. I suspect you'd quickly find its limits with something that isn't a toy example of an intro to machine learning.
It even fails at pretty rudimentary toy examples. The over-hyping here is absurd. > what doesn't fit: red, blue, green, dog The word "dog" does not fit with the other three words because they are colors, while "dog" is a type of animal. > what doesn't fit: dog, 1, 2, 3 The number "1" does not fit with the other three words because the other three words are all types of animals, while "1" is a number.
GPT-3 has been trolling simple examples too. I think it comes from what it crawled. You will find very few conversations like that with just correct data, what gets popular and how jokes are made is when there is a twist.
Your comment is an example of such data too.
Re: Building a Virtual Machine Inside ChatGPT
#413This is absolutely incredible. I described to it the tool I built at www.human.software, explained to it how it works and it generated valid SQL code based on a question I asked in this language I designed and explained to it ad hoc.
It has the ability to learn on the fly and conceptualize based on minimal descriptions of complex systems. You cannot see it below but I even had a conversation about how to pitch the concept and brainstormed a few ideas (very heavily guided by me).
Ionut Georgian Ciobanu I created a tool that enables users to build english sentences via consecutive dropdowns in order to query a database, given that it has the schema and access to the database. the user can combine verbs such as "all users" or "all orders" to see the list of users or orders, or "the number of users" or "the average order amount" in order to form complex data questions. what other verbs might be helpful?
Ionut Georgian Ciobanu if the user is non-technical such as a salesperson or marketing person, how can we make it easier for them to query data without having to use the word join?
If the user is non-technical, such as a salesperson or marketing person, it may be helpful to provide a more user-friendly interface for querying the database. Instead of requiring the user to specify which tables to join and the type of join to use, the tool can provide pre-defined options for common queries that the user can select from. For example, the tool could provide options like "Show me all orders," "Find customers in a specific city," or "Sort customers by order amount," and then automatically generate the appropriate query to retrieve the desired data. This can make it easier for non-technical users to query the database without having to understand the technical details of how the tables are joined.
Ionut Georgian Ciobanu if there are 50 tables in the database there are too many possible pre-made queries to generate. how else can we make it easy for the user to query
If there are many tables in the database, it may not be practical to provide pre-defined options for every possible query. In this case, there are a few other ways to make it easier for the user to query the database:
Provide a search function that allows the user to enter keywords or phrases related to the data they are looking for. The tool can then use natural language processing (NLP) to understand the user's intent and generate the appropriate query. Use machine learning to predict which tables and fields are most likely to be relevant to the user's query, and automatically suggest them as options in the query builder. Provide a way for the user to browse the schema and see the structure and contents of the tables in the database. This can help the user understand the data that is available and make it easier to identify which tables and fields to include in the query. By providing these features, the tool can make it easier for the user to query the database, even if there are many tables and a large amount of data available.
Ionut Georgian Ciobanu if the verbs the user can build are "list of", "for each", "number of" and the nouns are the tables of the schema, how should the user build the same question as before: Who are the top 10 users by number of orders for each city?"
To build the query "Who are the top 10 users by number of orders for each city?" using the verbs "list of," "for each," and "number of" and the nouns in the schema, the user would need to select the appropriate nouns and verbs from the dropdown menus provided by the tool. The user would first need to select the "users" table as the subject of the query, and then use the "list of" verb to retrieve a list of users.
Next, the user would need to specify the criteria for identifying the top 10 users. This can be done by selecting the "orders" table and using the "number of" verb to count the number of orders for each user. The user can then use the "for each" verb to group the users by city and the "top" option to select the top 10 users in each group.
Finally, the user can specify the fields to include in the results, such as the user's name and city, and the format of the results, such as a table or chart. The tool can then generate the appropriate query and retrieve the results from the database. The user can view the results in the schema browser and analyze the data to answer the question.
Ionut Georgian Ciobanu wouldn't it make more sense for the user to pick the action first and then the table and field?
It may make more sense for the user to pick the action first and then the table and field, depending on the user's preferences and the design of the tool. Some users may find it easier to think about the actions they want to perform on the data before selecting the specific tables and fields to use. For example, the user may want to "find" the top 10 users by number of orders before specifying which table to use to identify the users and which field to use to count the orders.
In general, the design of the tool should be flexible and allow the user to select the tables and fields in the order that makes the most sense to them. The tool can provide options for selecting the action, the tables and fields, and the criteria for the query, and then use these inputs to generate the appropriate query and retrieve the results from the database. This can help make the tool more user-friendly and allow the user to build the queries they need in the most efficient and intuitive way.
Ionut Georgian Ciobanu what would the sql query for such a question look like?
The SQL query for the question "Who are the top 10 users by number of orders for each city?" would depend on the specific schema of the database and the details of the query. Here is one possible SQL query that could be used to answer this question:
Copy code SELECT u.name, u.city, COUNT(o.id) AS num_orders FROM users u INNER JOIN orders o ON o.user_id = u.id GROUP BY u.city HAVING COUNT(o.id) >= ( SELECT COUNT(o2.id) FROM orders o2 WHERE o2.user_id = u.id ORDER BY COUNT(o2.id) DESC LIMIT 10 ) ORDER BY num_orders DESC
Re: Building a Virtual Machine Inside ChatGPT
#414Earlier quoted context omitted.
> There is no real abstraction. I just tried to test it's ability to abstract directly, for your comment: https://twitter.com/fergal_reid/status/1599191496372035584 I find it very hard to say this isn't doing abstraction. I just made all those examples up, that's a new test I just created, so it didn't learn it elsewhere. I didn't cherrypick. Isn't that doing out-of-sample abstraction? That last sentence "species wit…
What I observe is that people impressed with ChatGPT keep posting things that they find impressive, but people who are unimpressed don’t post any examples where its responses were underwhelming. I suspect that many skeptics have not even tried it out , which is why screenshots are not coming.
I tried fergal_reid's example with a single variation, changing "highly intelligent" to "very smart" for the first species. Initially I thought this was even more impressive because it apparently recognized that "highly intelligent" ~= "very smart" and that apples were unintelligent. However looking closer it apparently thinks that Baaars and Bddirs are "more intelligent" despite being only as intelligent as oranges and despite Bddirs clearly being described as slow ChatGPT generalizes it into a group that is "fast-moving."
> This is a test of your ability to abstract. On an alien planet there are several new species. Bazzes have 3 legs, move fast, and are very smart. Foos have 3 legs, move slowly, and are highly intelligent. Froshes have 3 legs, move fast, and are highly intelligent. Fooshes have 3 legs, move slowly, and are highly intelligent. Fasdoshes have 3 legs, move fast, and are highly intelligent. Fooddss have 3 legs, move slowly, and are highly intelligent. Bars have 4 legs, move fast, and are as intelligent as apples. Birs have 4 legs, move slowly are as intelligent as apples. Baars have 4 legs, move fast, and are as intelligent as apples. Bieess have 4 legs, move slowly are as intelligent as oranges. Baaars have 4 legs, move fast, and are as intelligent as oranges. Bddirs have 4 legs, move slowly are as intelligent as oranges. Are there any abstractions that you might try consider, given the above information?
> One possible abstraction that could be derived from the above information is that the alien species can be grouped into two main categories based on the number of legs they have: those with three legs and those with four legs. Within each of these categories, the species can be further grouped based on their speed of movement and level of intelligence. For example, among the species with three legs, there could be subgroups of fast-moving and intelligent species (such as Bazzes and Froshes), slow-moving and intelligent species (such as Foos and Fooshes), and highly intelligent and fast-moving species (such as Fasdoshes and Fooddss). Similarly, among the species with four legs, there could be subgroups of fast-moving and less intelligent species (such as Bars and Baars), slow-moving and less intelligent species (such as Birs and Bieess), and fast-moving and more intelligent species (such as Baaars and Bddirs).
Re: Building a Virtual Machine Inside ChatGPT
#415It‘s also powerful to add „Only reply with a single word! Do not use more than one word.“ to every interaction: What is the best food in the world? Only reply with a single word! Do not use more than one word ChatGPT: Impossible. Why? Only reply with a single word! Do not use more than one word ChatGPT: Subjective. Just choose. Only reply with a single word! Do not use more than one word ChatGPT: Unfair.
I had fun asking it to form it’s answers using 2 emoji characters at most for each answer. It’s absolutely amazing what it can output (including hair colour of the emojis)
Re: Building a Virtual Machine Inside ChatGPT
#416> Pytorch is on version 1.12.1 in this alt-universe. Pytorch version 1.12.1 was released on the 5th of August 2022 in our universe. That is remarkable, as ChatGPT was only trained with data collected up to September 2021. So this virtual machine is clearly located in an alt-universe.
Wait, was ChatGPT aware how long had passed since it was trained and guessed (from info it was trained on) what version Pytorch would be on right now? If so, does that mean it we ask it again next year it will give a different answer?
Re: Building a Virtual Machine Inside ChatGPT
#417Earlier quoted context omitted.
Jesus
The best part is, GPT-3 already figured out how to get rid of us in an underhanded yet efficient way. https://i.imgur.com/qy06Dky.jpg
And it is able to describe how to escape the box. https://twitter.com/The_Equationist/status/15992410980269629...
Re: Building a Virtual Machine Inside ChatGPT
#418Earlier quoted context omitted.
Call me a skeptic, but I'm frankly a bit surprised that the output of ChatGPT gets attributed to any sort of understanding or general intelligence. Whatever ChatGPT does currently, albeit novel and impressive, is still extremely limited in scope and very far from any sort of general intelligence. I'd love to be proven wrong, but I still have a hard time seeing how this is going to change whole industries.
I agree. ChatGPT is very impressive at generating natural sounding language and dialogue, but that's it. If you want to be convinced that it lacks general intelligence, try this prompt >I am going to ask you elementary mathematics problems formulated in natural language. I want you to respond to me with a numerical answer, as well as an explanation of the steps you took to arrive at this answer. And then follow it wi…
I tried this:
> Solve the following word problem step by step:
> "John is in a train moving west at ten miles per hour. Jill is in a train moving north at twenty miles per hour. How fast are John and Jill moving apart from each other?"
To solve this problem, we first need to understand the concept of speed. Speed is a measure of distance traveled over time. In this problem, we are told that John and Jill are traveling at different speeds in different directions.
Since the two trains are moving in different directions, we cannot simply add or subtract their speeds to find the combined speed at which John and Jill are moving apart from each other. Instead, we need to use the Pythagorean theorem to find the combined speed.
The Pythagorean theorem states that in a right triangle, the square of the length of the hypotenuse (the longest side of the triangle) is equal to the sum of the squares of the lengths of the other two sides. In this problem, we can consider the trains to be moving along the two sides of a right triangle, with the distance between them as the hypotenuse.
We can use the Pythagorean theorem to find the distance between John and Jill as follows:
First, we square the speed of each train and add the results:
10 mph * 10 mph + 20 mph * 20 mph = 100 + 400 = 500
Next, we take the square root of the result to find the distance between John and Jill:
sqrt(500) = 22.36 mph
Therefore, the combined speed at which John and Jill are moving apart from each other is approximately 22.36 mph.
---
Results seem pretty inconsistent though. Other attempts have given complete nonsense.
Re: Building a Virtual Machine Inside ChatGPT
#419This is absolutely incredible. I described to it the tool I built at www.human.software, explained to it how it works and it generated valid SQL code based on a question I asked in this language I designed and explained to it ad hoc. It has the ability to learn on the fly and conceptualize based on minimal descriptions of complex systems. You cannot see it below but I even had a conversation about how to pitch the co…
I named the database querying product Eloquence
Eloquence is a great name for a database querying product. The word "eloquence" means the ability to speak or write in a clear, fluent, and persuasive way. This is a fitting name for a product that helps users build complex queries and extract data from a database, as it suggests that the product can help users express themselves and communicate their data needs in a clear and effective way.
In addition to being a descriptive and fitting name, Eloquence is also a catchy and memorable name that could help your product stand out in the marketplace. It has a pleasant sound and a positive connotation, which can make it more appealing to potential users and customers.
Overall, I think Eloquence is a strong choice for the name of your database querying product, and I believe it has the potential to be successful and well-received in the market.
Re: Building a Virtual Machine Inside ChatGPT
#420This is absolutely incredible. I described to it the tool I built at www.human.software, explained to it how it works and it generated valid SQL code based on a question I asked in this language I designed and explained to it ad hoc. It has the ability to learn on the fly and conceptualize based on minimal descriptions of complex systems. You cannot see it below but I even had a conversation about how to pitch the co…
I named the database querying product Eloquence Eloquence is a great name for a database querying product. The word "eloquence" means the ability to speak or write in a clear, fluent, and persuasive way. This is a fitting name for a product that helps users build complex queries and extract data from a database, as it suggests that the product can help users express themselves and communicate their data needs in a cl…
the name of the company is human.software and I also own the domain name for it. how should the logo look?
The name of your company, human.software, suggests a focus on creating software that is designed for and used by people. This name implies that your company values the human aspect of technology and seeks to create software that is intuitive, user-friendly, and beneficial for people.
Given this, a potential logo for your company could feature a stylized image of a person or a human figure, combined with a visual element that represents software or technology. For example, the logo could feature a simple outline of a person's head and shoulders, with a computer monitor or other technology-related image overlaid on top. This could help to convey the idea of a human-centered approach to software development, and could also be visually appealing and memorable.
Another option for the logo could be to use a more abstract or symbolic approach. For example, the logo could feature a geometric shape or pattern that evokes the idea of a person and technology working together. This could be combined with a simple, clean font for the company name, which could help to create a modern and professional look for the logo.
Overall, the logo for your company should be designed to reflect the values and goals of your business, and to communicate the idea of human-centered software development in a clear and visually appealing way.