Manifold Community Site: AI and SQL

RE: AI and SQL

lionel — Fri, 24 Jan 2025 16:59:47 GMT

Manifold has developed its own engine algorithm, suggesting they created their AI engine from scratch, which requires powerful APU(GPU +CPU) and NPU. However, the use of SQL in Manifold GIS is unique to their system because of : functions name and parameters, type safe query system, script, way to import/include/call .. libraries/Script/function.

Given the manifold developers' experience with databases handling textual (ASCII -ISO), raster (area of points, color, transparency) and vector (line of points, color, transparency) data, the next logical step was to use new AI-related databases and tools.

-the use of Retrieval Augmented Generation (RAG) in AI is particularly useful when working with SQL databases.

-Natural Language Processing (NLP) plays a crucial role in bridging the gap between human language and structured query language (SQLL) , making database interactions more accessible and efficient

---------------------------------Some AI SQL web implementation ------------------------------------------

In the context of SQL and AI , the competitors use AI base on many database model storage :

-chapGPT : NoSQL ( chat history, user prompts) , Vector ( learn search),NLP

ChatGPT and other large language models: While not specifically designed for SQL, these AI tools can be used to generate and optimize complex SQL queries when provided with the appropriate context and schema information.

GitHub - billmei/every-chatgpt-gui

-Sequel : Vector, Relational (schema understanding), NLP, Query optimization

It offers a streamlined experience by automatically fixing SQL errors and generating visual representations of query results

Sequel | AI SQL Generator and AI Data Analyst

-SQLAI.ai : Vector, Relational ( schema import and autosuggest, automatic table preselection, batch mode index, RAG)

SQLAI.ai: This platform offers features like SQL query generation, optimization, and syntax validation. It can handle complex queries and large schemas, making it suitable for advanced SQL code writing.

Generate SQL Queries in Seconds for Free - SQLAI.ai

-Text2SQL.ai : Vector, Relational,NoSQL

Text2SQL.ai: This tool can generate optimized SQL queries for various database systems, including MySQL, PostgreSQL, Oracle, and more.It supports complex queries with multiple tables and can handle large database schemas with over 600 tables.

Text2SQL.ai | Text to SQL & AI Query Generator

-Blaze SSL AI : Vector, Relational, NoSQL,Graph, Time-Series.Understand metadata,schema, NLP ( english to SQL) and optimize Query .

Specializes in database development with built-in domain knowledge. It can generate SQL queries, explain complex queries, and optimize performance

Blaze SQL AI: This AI Data Analyst does your work in seconds

-AI2SQL

AI2sql: This AI-driven SQL query generator allows users to input instructions in natural language to create SQL queries, which can be helpful for complex scenarios.

SQL Query Builder & Generator - AI Powered Database Assistant

-DataGrip AI Assistant:

Offers natural language query generation, SQL explanation, and optimization. It is a database management tool by JetBrains ( not AI) .

DataGrip: The Cross-Platform IDE for Databases & SQL by JetBrains

-SQLChat

SQL Chat is a chat-based SQL client and editor that uses AI to simplify database interactions.It supports multiple database platforms, including MySQL, PostgreSQL, and SQL Server

SQLChat 5:41:28 PM

-outerbase

Users can ask questions in plain English, and the AI translates these natural language inputs into SQL queries.. This functionality allows users to interact with their databases without needing to write complex SQL code, making data exploration more accessible to those without extensive SQL knowledge

Outerbase | The interface for your database

It's important to note that while SQL in AI can provide creative solutions, human oversight and validation remain crucial to ensure accuracy and appropriateness in database operations.

post using my daily AI free tool call perplexity :AI focuses on providing accurate, up-to-date information from the web. Perplexity provides citations and references for its information, allowing users to verify sources.

RE: AI and SQL

lionel — Fri, 24 Jan 2025 20:38:12 GMT

many databases implement functionnalities for AI : in it own core using module or access to tools ( LangChain LC , LlamaIndex LI , Semantic Kernel SK, Atlas from Nomic AI, PeerAI From PeerIsland , pureInsights) or access to external AI platform like Azure OpenAI (AO), Amazon Bedrock (AB) , Google cloud Vertex AI ( GCV) or agent that have access to AI Services like SKySQL, Klu LLM platform, Aerospike, Couchbase Capella ,Together AI !

Microsoft |Azur SQL .....|Vector search |AI service AO - Copilot -Fabric- AI tools LC , SK

timescale....| PostgresQL | PostgresML |pgai pgvector | AI services ( AO, GCV)

investor | MongoDB | MongoDB AI -Meta AI -MAAP- Atlas | AI Services AB,

Redis Labs |Redis |Redis AI module -vector index -RAG- tensorFlow PyTorch- AI Tools + services

Machine Learning (ML) is a subset of AI that focuses on algorithms allowing computers to learn from and make predictions based on data. It involves training models on data to identify patterns and make decisions without being explicitly programmed.Deep Learning (DL) is a subset of Machine Learning.Specific chips are use for ML.

Several AI and machine learning frameworks support : Linear Regression,Logistic Regression ,Decision Trees ,Matrix Operations like Scikit-learn, TensorFlow,PyTorch, RapidMiner,H2O.ai .

Artificial Intelligence (AI) is a broad field that encompasses any technique enabling computers to mimic human intelligence. This includes tasks like problem-solving, learning, and pattern recognition.ML involves training algorithms on data to make predictions or decisions without being explicitly programmed for specific tasks. ML aims to create models that can learn from and make decisions based on data.

Natural Language Processing (NLP) is a subfield of AI that focuses on the interaction between computers and human language. It aims to enable machines to understand, interpret, and generate human language in a way that's both meaningful and useful. NLP encompasses tasks like speech recognition, sentiment analysis, language translation, and text summarization.The tools are : Hugging Face Transformers, Google Cloud Natural Language, LTK (Natural Language Toolkit), SpaCy, Stanford CoreNLP,IBM Watson Natural Language ,Amazon Comprehend, AllenNLP, Lexalytics, Clarabridge.LLMs are subset of NLP ,representing a more advanced and specialized application of language processing techniques

A Large Language Model (L-LM) is a type of AI model designed tounderstand and generate human language. These models are trained on vast amounts of text data and can perform a variety of natural language processing tasks, such as text generation, translation, summarization, and question answering.

The big name are : GAFAM : Gemini & PaLm ( Google use Trilium TPU or Axion Arm cpu) , Llama (Meta ), GPT ( Microsoft use OpenAI technology), titan/Nova/Olympus ( Amazon use AnnapurnaLabs graviton Arm licence for AWS ) , GTP and "noname" (Apple has ARM licence and use Google TPU ) outside GAFA there is GPT (OpenAI), Claude (Anthropic).For communication dialogue base on AI there is LaMDA ,Claude 3 haidu , Qwen, RoBERT (MAsked M-LM) base on BERT and T5, Meena and BlenderBot, Replika,chatGPT.

The AI is also a battle of CPU Architecture on power management . RISC-V , intel x86 can't compete against ARM that licence the right to design CPU or to use ARM architecture cortex CPU design by team in Austin Texas, Sophia Antipolis, cambridge.

custom ARM => Apple (M4), Amazon Annapurna Labs ( graviton ) ,NVIDIA( tegra , Grace , Blackwell with mediatek is not ARM base but NVidia ), qualcomm (orion with help of nuva not ) ,Samsung (exynos) , Huawei ( voir HiSilicon), HiSilicon .

cortex ARM integration => ARM Holdings is the owner and designer of All Cortex familly licenced to many company : NVIDAI 5 Jetson orin/wavier/nano/AGX ) Qualcomm (snapdragon), Broadcom ( BCM dans Raspberry Pi et Compute Module), HiSilicon (kirin , kunpeng) , AWS ( Graviton) Texas intrument (JAcinto 7) , Mediatek ( Dimensity) , Rockchip (RK chip) , NXP (i.MX, LS LX series) , Marvell ( armada dans CCAudio), AMD , Intel (XScale Marvell), mediatek.

The foundry that deliver manufacture optimized system on chip ( SoC) for ARM Cortex processor design are : intel Foundry Services ( IFS), Samsung Factory and perhaps , Taiwan Semiconductor Manufacturing Company (TSMC), GlobalFoundries.

RE: AI and SQL

Dimitri — Sat, 25 Jan 2025 10:00:42 GMT

Don't forget the current biggest competitor to ChatGPT: DeepSeek - significantly better at code and math than ChatGPT 4o at all the usual benchmarks, and better than ChatGPT in English (...not bad for a chinese AI...) in seven out of nine benchmarks.

RE: AI and SQL

lionel — Mon, 27 Jan 2025 20:03:29 GMT

the world of AI is vast ... thank's a lot for the link

Top Apps & Games for iPhone on the iOS App Store in the United States · Appfigures

RE: AI and SQL

lionel — Tue, 05 Aug 2025 20:44:29 GMT

Query using LLM is very different than query using relationnal database like manifold. So the concept seem to be all about vector, embedding, graph, semantic concepts.

So there are many differents databases structures from manifold 8 9 and 10 ( 10 with AI ).

If manifold train their model heavely; the end user don't need Super calculator even if new data should be supported in real time (RAG) from the heavy train LL database model !!

******************SQL AI LLM ******************************

general : GPT

general and then train SQL = pre train

--------------LLM => SQLAI ( GPT 4) , Text2SQL.ai (Google T5), OuterBase ( EZQL )

--------------unknown ( LLM open AI ?) => AI2SQL , DataGrip AI SQLChat

--------------SSL => Blaze SSL AI (GPT 3) :

specific train SQL => Sequel ,

***************************************************

Does manifold ( LLM ?) has been trained from scratch only on SQL code and query data ?

Does Manifold use CUDA for train their own LLM ?

thank's

link rumor for 24 GB RTX 5080 GDDR7 .....

https://pro.arcgis.com/en/pro-app/3.4/help/analysis/ai/geoai.htm

RE: AI and SQL

lionel — Tue, 05 Aug 2025 21:08:46 GMT

https://github.com/AGI-GIS/BB-GeoGPT

https://encord.com/blog/segment-anything-model-explained/

RE: AI and SQL

lionel — Sun, 10 Aug 2025 12:12:33 GMT

I am concerned that general AI does not require specialised training to reach the level of older AI with specialised training.

If this is the case, paying for a licence to use general AI and learning how to use it will be more cost-effective than subscribing to several specialised AI systems.

https://www.oneusefulthing.org/p/which-ai-to-use-now-an-updated-opinionated

https://www.oneusefulthing.org/p/using-ai-right-now-a-quick-guide