Alignment (Alignment)
Alignment
Dwumadi a ɛma AI nhyehyɛe bi atirimpɔw, nea ɛma fi, ne ne suban ne nnipa atirimpɔw ne gyinapɛn hyia. Eyi ho hia titiriw wɔ nhyehyɛe a ɛkɔ anim a ebetumi anya suban a wɔamfa nkyerɛkyerɛ mu pɛpɛɛpɛ no mu.
Nhwɛso: Sɛ wode bɔhyɛ ma sɛ chatbot a ɛboa adwenemhaw ho no remfa nneɛma bɔne nkyerɛ da, sɛnea nsɛmmisa no te biara.
Application Programming Interface (API) (Application Programming Interface (API))
Application Programming Interface (API)
Mmara ne nhyehyɛe a wɔahyɛ da a ɛma software nhyehyɛe ahorow betumi abom akasa na wɔakyere data.
Nhwɛso: Sɛ wode OpenAI API di dwuma de soma nsɛmmisa na wunya kasa nhwɛso a wɔayɛ no wɔ wo web app mu.
Artificial General Intelligence (AGI) (Artificial General Intelligence (AGI))
Artificial General Intelligence (AGI)
AI kwan bi a É›wÉ” adwene mu a ebetumi ayÉ› adwuma biara a nnipa betumi ayÉ›. Æma nhomasua no trÉ›w kÉ” mmeae ahorow so.
Nhwɛso: AGI nhyehyɛe betumi asua nnwom a wɔyɛ, ayɛ oprehyɛn, na watumi apasa nyansapɛ sɔhwɛ a ɛho nhia sɛ wɔyɛ adwuma pɔtee bi ho nhyehyɛe.
Artificial Intelligence (AI) (Artificial Intelligence (AI))
Artificial Intelligence (AI)
Nnipa adwene a wɔde ayɛ wɔ mfidie mu a wɔahyɛ ho nhyehyɛe sɛ ɛbɛdwene, adwene, na ayɛ adwuma wɔ ne ho so.
Nhwɛso: AI na ɛma ankorankoro boafo te sɛ Siri ne kar a ɛkɔ ne ho te sɛ Tesla Autopilot yɛ adwuma.
AI Ethics (AI Ethics)
AI Ethics
Nkyerɛkyerɛmu a ɛfa AI nkɔso ne ne dwumadi ho, a ɛka pɛyɛ, ankorankoro nsɛm, akontabuo, ne nyiyim a enni mu ho.
Nhwɛso: Mmara a wɔyɛ de siw akatua ho nhyehyɛe a ɛma wɔyɛ nyiyim wɔ ɔbarima ne ɔbea nna anaa mmusuakuw ho no ano.
Augmented Intelligence (Augmented Intelligence)
Augmented Intelligence
Nkabom nhwɛso a AI boa na ɛma nnipa adwene yɛ kɛse sen sɛ ɛbɛsesa no.
Nhwɛso: AI-powered radiology nnwinnade a ɛkyerɛ nsɛm a ɛnteɛ kyerɛ adɔkotafo, a wɔyɛ gyinae a etwa to.
Autonomous Agent (Autonomous Agent)
Autonomous Agent
AI nhyehyɛe a ebetumi ayɛ ne gyinae ankasa na ayɛ nneɛma de adu ne botae ho a nnipa mmu ne ho.
Nhwɛso: Kar a ɛde ne ho kɔ de nneɛma kɔma a ɛfa kwan so na ɛkwati akwanside ahorow wɔ ne ho so.
Backpropagation (Backpropagation)
Backpropagation
Akwan a wɔfa so kyerɛkyerɛ neural networks denam nneɛma a wɔde kyerɛ no a wɔsesa no fi nea ɛma fi kosi nea wɔde hyɛ mu no so, na ɛma mfomso a wɔhyɛ ho nkɔm no so tew.
Nhwɛso: Wɔde di dwuma wɔ mfonini a wɔde kyerɛkyerɛ mu de tew mfomso a ɛba wɔ nkyerɛwde a wɔde nsa akyerɛw no mu no so.
Bias (Algorithmic Bias) (Bias (Algorithmic Bias))
Bias (Algorithmic Bias)
AI nea ɛma fi a ɛnteɛ na ɛyɛ nhyehyɛe a ɛma wɔpɛ nneɛma bi sen nneɛma afoforo esiane data a wɔde kyerɛkyerɛ mu a ɛnteɛ anaa ɛnkyerɛkyerɛ mu yiye nti.
Nhwɛso: Anim a wɔde hu nnipa nhyehyɛe a ɛtaa hu nnipa a wɔyɛ kɔlafo mfomso esiane sɛ wɔn ananmusifo nni hɔ wɔ nhomasua data mu nti.
Big Data (Big Data)
Big Data
Data kɛse a ɛhwehwɛ nnwinnade pɔtee de kora, hwehwɛ mu, na woyi mfaso fi mu, wɔtaa de di dwuma de kyerɛkyerɛ AI nhwɛso ahorow.
Nhwɛso: Sɛ wode ɔpepem pii a nnipa ne wɔn ntam nkitahodi di dwuma de kyerɛkyerɛ nneɛma a wɔde ma ho nhyehyɛe ahorow ma e-commerce platforms.
Black Box Model (Black Box Model)
Black Box Model
AI anaa machine learning nhwɛso bi a ne mu adwene mu yɛ den sɛ nnipa bɛkyerɛkyerɛ mu, na ɛma ɛyɛ den sɛ wubehu sɛnea wɔyɛ gyinae.
Nhwɛso: Deep neural network a wɔde di dwuma de gye bosea nanso ɛmfa nkyerɛkyerɛmu a emu da hɔ mma nea enti a wɔagye obi na wɔapo obi foforo.
Cognitive Computing (Cognitive Computing)
Cognitive Computing
AI nhyehyɛe a wɔayɛ no sɛ ɛbɛyɛ nnipa adwene mu nneɛma, te sɛ adwene ne nhomasua, denam akwan te sɛ NLP ne pattern recognition so.
Nhwɛso: Cognitive computing nhyehyɛe a ɛboa mmara ho adwumayɛfo ma wɔhwehwɛ mmara mu nsɛm na wɔhyɛ nea ɛbɛba ho nkɔm.
Computer Vision (Computer Vision)
Computer Vision
Adwene a wɔde ayɛ (AI) mu adwuma bi a ɛma kɔmputa tumi kyerɛkyerɛ na ɛyɛ mfonini ne video data.
Nhwɛso: Anim a wɔde hu nnipa nhyehyɛe a ɛhu nnipa wɔ ahobammɔ mfonini mu denam computer vision so.
Corpus (Corpus)
Corpus
Nkyerɛwde anaa kasa kɛse a wɔde di dwuma de kyerɛkyerɛ kasa nhwɛso ahorow.
Nhwɛso: Common Crawl data set yɛ web corpus a wɔde di dwuma de kyerɛkyerɛ kasa nhwɛso akɛse te sɛ GPT.
Data Drift (Data Drift)
Data Drift
Nsɛm a wɔde hyɛ mu no sesa bere kɔ so, na ɛma nhwɛso no yɛ adwuma bɔne.
Nhwɛso: Nhwɛso a ɛhyɛ nkɔm a ɛfa mfidie ho no yɛ adwuma bɔne bere a wɔde sensor foforo ba no.
Data Labelling (Data Labelling)
Data Labelling
Dwumadi a wɔde tags anaa labels kyerɛ data ho de ma ɛfata sɛ wɔde di dwuma wɔ nhomasua a wɔhwɛ so mu.
Nhwɛso: Sɛ wode akatua pii a ɛfa kokoram ho kyerɛ sɛ ɛyɛ papa anaa ɛyɛ bɔne de kyerɛkyerɛ kokoram a wɔde hu no mu.
Data Mining (Data Mining)
Data Mining
Dwumadi a wɔde hu nneɛma a ɛho hia, nkitahodi, ne nneɛma a ɛnteɛ wɔ data kɛse mu.
Nhwɛso: Retailers a wɔde data mining di dwuma de hu sɛ nnipa a wɔtɔ nappies no taa tɔ beer nso.
Deep Learning (Deep Learning)
Deep Learning
Machine learning mu adwuma bi a ɛde multi-layered neural networks di dwuma de yɛ data mu nneɛma a emu yɛ den.
Nhwɛso: Wɔde deep learning di dwuma wɔ kasa nhwɛso ahorow te sɛ GPT-4 ne mfonini a wɔde yɛ nhwɛso ahorow te sɛ Stable Diffusion.
Diffusion Models (Diffusion Models)
Diffusion Models
Generative models kuw bi a wosua sɛnea wɔyɛ data denam random noise a wɔsesa no kɔ outputs a wɔahyɛ ho nhyehyɛe mu.
Nhwɛso: Stable Diffusion yɛ mfonini a ɛyɛ nokware fi text prompts mu denam diffusion akwan so.
Embedding (Embedding)
Embedding
Data a wɔde kyerɛw no wɔ nɔma mu, wɔtaa de di dwuma de kyerɛ nsɛmfua, mfonini, anaa nsɛm a ɛwɔ nkyerɛase.
Nhwɛso: Wɔ NLP mu no, asɛmfua 'bank' no betumi anya embeddings a ɛte sɛ 'money' nanso ɛsono 'riverbank' a ɛfa ho no.
Epoch (Epoch)
Epoch
Dwumadi a wɔde kyerɛkyerɛ machine learning nhwɛso bi mu no nyinaa.
Nhwɛso: Sɛ data set bi wɔ nhwɛso 1,000 na nhwɛso bi hu ne nyinaa pɛnkoro wɔ nhomasua mu a, ɛno ne epoch biako.
Ethical AI (Ethical AI)
Ethical AI
Nhyehyɛe ne dwumadi ho nyansapɛ a ɛma AI mfiridwuma yɛ adwuma wɔ ɔkwan a ɛda hɔ, ɛyɛ pɛ, na ɛne ɔmanfo gyinapɛn hyia.
Nhwɛso: AI a wɔde bɔhyɛ ma a ɛka bias checks de siw nyiyim a wɔyɛ tia nnipa kakraa bi no ano.
Expert System (Expert System)
Expert System
AI nhyehyɛe a ɛyɛ nnipa ɔbenfo bi gyinae a wɔyɛ no wɔ ɔkwan pɔtee bi so denam mmara ne adwene so.
Nhwɛso: Expert system a wɔde di dwuma wɔ kuayɛ mu de kyerɛkyerɛ nnɔbae ho nneɛma a ɛfa asase data ne mmoawa ho.
Explainable AI (XAI) (Explainable AI (XAI))
Explainable AI (XAI)
AI nhyehyɛe a wɔayɛ no sɛ ɛbɛma ne mu dwumadi ne ne gyinae no mu ada hɔ ama nnipa, na ɛma wɔn ahotoso ne akontabuo no kɔ soro.
Nhwɛso: Aduruyɛ mu AI a ɛkyerɛkyerɛ mu sɛnea ɛyɛ adwuma na ɛkyerɛkyerɛ mu sɛnea ɛyɛ gyinae.
Few-shot Learning (Few-shot Learning)
Few-shot Learning
Machine learning kwan bi a wɔde kyerɛkyerɛ nhwɛso bi anaa wɔyɛ no yiye denam nhwɛso kakraa bi a wɔde kyerɛw no so.
Nhwɛso: Sɛ wode LLM yɛ mmara ho email bere a woakyerɛ no nhwɛso 10 pɛ.
Fine-tuning (Fine-tuning)
Fine-tuning
Dwumadi a wɔde fa nhwɛso a wɔakyerɛkyerɛ mu dedaw na wɔsan kyerɛkyerɛ mu wɔ data set foforo, ketewaa bi so de yɛ adwuma pɔtee bi.
Nhwɛso: Sɛ wode LLM a ɛyɛ foforo te sɛ GPT yɛ mmara ho nkrataa a wɔde kyerɛw no.
Foundation Model (Foundation Model)
Foundation Model
Nhwɛso kɛse a wɔakyerɛkyerɛ mu wɔ data ahorow so a wobetumi de ayɛ adwuma ahorow pii.
Nhwɛso: GPT-4 ne PaLM 2 yɛ nhwɛso ahorow a ebetumi atumi ayɛ nsɛm a wɔka tiawa, nsɛmmisa ne mmuae, nkyerɛase, ne nea ɛkeka ho.
Fuzzy Logic (Fuzzy Logic)
Fuzzy Logic
Adwene kwan bi a ɛfa nneɛma a ɛbɛn ho sen sɛ ɛbɛyɛ nokware/atoro (binary) adwene, ɛboa ma wɔyɛ adwene wɔ nea ɛnteɛ mu.
Nhwɛso: Wɔde di dwuma wɔ wim tebea ho nhyehyɛe ahorow mu de sesa ɔhyew a ɛfa nneɛma a ɛyɛ den te sɛ 'ɛyɛ hyew kakra' anaa 'ɛyɛ nwini kɛse'.
Generative Adversarial Network (GAN) (Generative Adversarial Network (GAN))
Generative Adversarial Network (GAN)
Generative model architecture bi a networks abien — generator ne discriminator — di asie de ma output no yɛ papa.
Nhwɛso: Wɔde GANs di dwuma de yɛ deepfake video anaa wɔyɛ mfonini a ɛyɛ nokware fi sketches mu.
Generative AI (Generative AI)
Generative AI
Adwene a wɔde ayɛ (AI) kuw bi a ebetumi ayɛ nsɛm foforo — te sɛ text, mfonini, nnwom, anaa video — fi nhomasua data mu.
Nhwɛso: ChatGPT a ɛyɛ blog posts anaa Midjourney a ɛyɛ digital artwork fi textual prompts mu.
Generative Pre-trained Transformer (GPT) (Generative Pre-trained Transformer (GPT))
Generative Pre-trained Transformer (GPT)
Kasa nhwɛso akɛse a OpenAI yɛe a ɛde transformer architecture di dwuma na wɔakyerɛkyerɛ mu wɔ text data pii so de yɛ kasa adwuma ahorow.
Nhwɛso: GPT-4 betumi atumi akyerɛw essays, akyerɛ kasa ase, na ayɛ nsɛm a wɔka tiawa a ɛho nhia sɛ wɔyɛ nsɛmmisa pii.
Genetic Algorithm (Genetic Algorithm)
Genetic Algorithm
Optimisation kwan bi a ɛfa abɔde mu nhyehyɛe ho a solutions no kɔ so denam mutation, crossover, ne selection so.
Nhwɛso: Wɔde di dwuma de yɛ neural network architectures a ɛyɛ adwuma yiye denam survival of the fittest a wɔyɛ no so.
Hallucination (Hallucination)
Hallucination
Nsɛm a ɛyɛ nokware nanso ɛnteɛ anaa ɛnteɛ a AI nhwɛso bi yɛ.
Nhwɛso: Kasa nhwɛso bi a ɛyɛ citation a enni hɔ anaa ɛma abakɔsɛm mu nsɛm a ɛnteɛ.
Heuristic (Heuristic)
Heuristic
Ɔkwan a ɛyɛ adwuma de siesie nsɛm a ɛnhyɛ bɔ sɛ ɛbɛyɛ solution a ɛyɛ pɛ nanso ɛfata ma botae a ɛwɔ hɔ no.
Nhwɛso: Sɛ wode rule of thumb di dwuma de bu delivery time wɔ logistics AI nhyehyɛe mu.
Hyperparameter (Hyperparameter)
Hyperparameter
Configuration value bi a wɔahyɛ no dedaw ansa na wɔakyerɛkyerɛ machine learning nhwɛso bi mu, te sɛ learning rate anaa layers dodow.
Nhwɛso: Sɛ wode batch size no sesa fi 32 kosi 128 de ma nhomasua no yɛ ntɛm na nhwɛso no yɛ adwuma yiye.
Inference (Inference)
Inference
Dwumadi a wɔde di dwuma wɔ machine learning nhwɛso a wɔakyerɛkyerɛ mu no de yɛ nkɔmhyɛ anaa wɔyɛ outputs fi input data foforo mu.
Nhwɛso: Sɛ wode GPT nhwɛso a wɔayɛ no yiye di dwuma de yɛ email ma customer support team.
Intent Detection (Intent Detection)
Intent Detection
Adwuma bi a ɛwɔ kasa a ɛyɛ abɔde mu ntease mu a nhyehyɛe no hu nea ɔde di dwuma no botae anaa atirimpɔw wɔ nsɛm mu.
Nhwɛso: Wɔ chatbot mu no, sɛ wuhu 'Mepɛ sɛ mekyere wimhyɛn' sɛ travel booking intent.
Internet of Things (IoT) (Internet of Things (IoT))
Internet of Things (IoT)
Mfidie a wɔde abom a ɛwɔ sensors, software, ne mfiridwuma afoforo a wɔde kora na wɔkyere data.
Nhwɛso: Smart thermostats ne fridges a ɛkyerɛkyerɛ dwumadi data na ɛsesa settings denam AI analytics so.
Interpretability (Interpretability)
Interpretability
Sɛnea nnipa betumi ate machine learning nhwɛso bi mu dwumadi ne ne gyinae ase.
Nhwɛso: Decision tree no mu da hɔ sen deep neural network efisɛ ne gyinae no betumi ahu.
Jupyter Notebook (Jupyter Notebook)
Jupyter Notebook
Open-source interactive computing environment a ɛma nnipa betumi akyerɛw code, ahu outputs, na wɔakyerɛw analysis wɔ interface biako mu.
Nhwɛso: Data scientists de Jupyter Notebooks di dwuma de yɛ machine learning models na wɔkyere results.
K-Nearest Neighbours (KNN) (K-Nearest Neighbours (KNN))
K-Nearest Neighbours (KNN)
Machine learning algorithm a É›yÉ› mmerÉ›w, non-parametric a wÉ”de di dwuma de yÉ› classification ne regression. ÆyÉ› gyinae denam nhwÉ›so a É›bÉ›n no so wÉ” feature space mu.
Nhwɛso: Sɛ wode nnua foforo kyerɛkyerɛ mu sɛ ɛyɛ apple anaa pear a, KNN hwehwɛ nnua a wɔakyerɛw no a ɛbɛn no wɔ ne kwan ne ne kɔla mu.
Knowledge Graph (Knowledge Graph)
Knowledge Graph
Data structure bi a ɛde nodes ne edges di dwuma de kyerɛkyerɛ na ɛkora nneɛma ne wɔn ntam nkitahodi.
Nhwɛso: Google's knowledge panel no yɛ adwuma denam knowledge graph a ɛka nneɛma te sɛ nnipa, mmeae, ne nsɛm.
Language Learning Model Optimisation (LLMO) (Language Learning Model Optimisation (LLMO))
Language Learning Model Optimisation (LLMO)
Akwan a wɔde di dwuma de ma kasa nhwɛso akɛse no yɛ adwuma yiye, ɛyɛ adwuma yiye, anaa ɛfata ma adwuma anaa mmeae pɔtee bi.
Nhwɛso: Sɛ wode quantisation ne instruction tuning di dwuma de yɛ LLM ma enterprise use.
Large Language Model (LLM) (Large Language Model (LLM))
Large Language Model (LLM)
Deep learning model bi a wɔakyerɛkyerɛ mu wɔ textual data pii so a ebetumi ayɛ, ate ase, na adwene wɔ nnipa kasa mu.
Nhwɛso: ChatGPT ne Claude yɛ LLMs a wɔakyerɛkyerɛ mu de boa wɔ nkyerɛwde, coding, ne nsɛmmisa ho.
Latent Space (Latent Space)
Latent Space
Nkyerɛkyerɛmu a ɛwɔ soro a ɛma inputs a ɛte sɛ nea ɛwɔ hɔ no bom, wɔde di dwuma wɔ generative models ne embeddings mu.
Nhwɛso: Wɔ mfonini a wɔde yɛ mu no, sɛ wode latent space di dwuma no betumi asesa nneɛma te sɛ hann anaa adwene.
Learning Rate (Learning Rate)
Learning Rate
Hyperparameter a ɛho hia wɔ nhomasua mu a ɛhwɛ sɛnea model weights no sesa wɔ loss gradient ho.
Nhwɛso: Learning rate a ɛkɔ soro no betumi ama wɔayɛ minima a ɛboro so, bere a rate a ɛba fam no ma nhomasua no yɛ brɛoo.
Machine Learning (ML) (Machine Learning (ML))
Machine Learning (ML)
AI mu adwuma bi a ɛma nhyehyɛe ahorow betumi asua ade fi data mu na wɔama wɔn dwumadi ayɛ papa a wɔamfa nhyehyɛe pɔtee bi nkyerɛkyerɛ mu.
Nhwɛso: Spam filters de machine learning di dwuma de kyerɛkyerɛ email sɛ ɛyɛ spam anaa ɛnyɛ spam denam nhwɛso a atwam so.
Model Drift (Model Drift)
Model Drift
Nsɛm a ɛba bere a nhwɛso bi yɛ adwuma bɔne bere kɔ so esiane data anaa wim tebea mu nsakrae nti.
Nhwɛso: Fraud detection model bi yɛ adwuma bɔne bere a fraud akwan no kɔ so no.
Model Training (Model Training)
Model Training
Dwumadi a wɔde data ma machine learning nhwɛso bi na wɔsesa ne parameters de tew mfomso so.
Nhwɛso: Sɛ wode recommendation engine kyerɛkyerɛ mu wɔ customer purchase history so de kyerɛ nneɛma foforo.
Multimodal AI (Multimodal AI)
Multimodal AI
AI nhyehyɛe a ebetumi ayɛ na abom data ahorow pii te sɛ text, mfonini, audio, ne video.
Nhwɛso: Nhwɛso bi te sɛ GPT-4 Vision a ebetumi akenkan text na akyerɛkyerɛ mfonini ase wɔ bere koro mu.
Natural Language Processing (NLP) (Natural Language Processing (NLP))
Natural Language Processing (NLP)
AI mu adwuma bi a É›fa kÉ”mputa ne nnipa (abÉ”de mu) kasa ntam nkitahodi ho. Æma mfidie betumi akenkan, ate ase, na wÉ”ayÉ› mmuae wÉ” nnipa kasa mu.
Nhwɛso: Wɔde NLP di dwuma wɔ voice assistants, language translation apps, ne chatbots.
Neural Network (Neural Network)
Neural Network
Machine learning nhwɛso bi a ɛfa nnipa adwene ho, a ɛwɔ nodes (neurons) a wɔde abom.
Nhwɛso: Neural networks na ɛwɔ deep learning models a wɔde di dwuma wɔ mfonini ne kasa a wɔde hu no mu.
Noise (Noise)
Noise
Nsɛm a ɛyɛ random anaa ɛho nhia wɔ data mu a ebetumi asiw nneɛma a ɛho hia ano na ama nhwɛso no yɛ adwuma bɔne.
Nhwɛso: Sensor mfomso anaa typo-filled data entries no betumi abu no sɛ noise.
Ontology (Ontology)
Ontology
Nhyehyɛe a wɔahyɛ ho nhyehyɛe a ɛkyerɛkyerɛ na ɛkyerɛkyerɛ nneɛma ntam nkitahodi wɔ ɔkwan bi so, wɔtaa de di dwuma wɔ semantic AI nhyehyɛe ahorow mu.
Nhwɛso: Ontology wɔ akwahosan mu no betumi akyerɛkyerɛ sɛnea nsɛm a ɛfa yare ho no ne nyarewa ne nnuru ho.
Overfitting (Overfitting)
Overfitting
Modelling mfomso bi a machine learning nhwɛso bi kora noise wɔ nhomasua data mu na ɛyɛ adwuma bɔne wɔ data foforo so.
Nhwɛso: Nhwɛso bi a ɛkora nhomasua mmuae nanso ɛntumi nni test data a wonhuu no ho dwuma no yɛ overfitted.
Predictive Analytics (Predictive Analytics)
Predictive Analytics
Sɛ wode data, algorithms, ne AI di dwuma de hu nea ɛbɛba daakye denam abakɔsɛm data so.
Nhwɛso: Retailers de predictive analytics di dwuma de hyɛ nneɛma bi ho nkɔm.
Pre-training (Pre-training)
Pre-training
Dwumadi a wɔde kyerɛkyerɛ nhwɛso bi mu wɔ data kɛse, foforo bi so ansa na wɔayɛ no yiye ama adwuma pɔtee bi.
Nhwɛso: GPT models no wɔakyerɛkyerɛ mu wɔ corpora akɛse so ansa na wɔayɛ no sɛ customer service chatbots.
Prompt Engineering (Prompt Engineering)
Prompt Engineering
Adwinni ne nyansahu a ɛfa prompts a ɛyɛ adwuma yiye a wɔde yɛ de kyerɛ kasa nhwɛso akɛse no output.
Nhwɛso: Sɛ wode system instructions te sɛ 'Reply as a polite tutor' ka ho no yɛ prompt engineering nhwɛso.
Quantisation (Quantisation)
Quantisation
Model compression kwan bi a ɛtew bits dodow a wɔde kyerɛ weights ne activations no so, na ɛma ɛyɛ adwuma yiye.
Nhwɛso: Sɛ wode model bi yɛ 32-bit kosi 8-bit no ma ɛyɛ adwuma yiye wɔ mobile devices so.
Quantum Computing (Quantum Computing)
Quantum Computing
Computing kwan foforo a ɛfa quantum mechanics ho, a ɛwɔ tumi a ɛyɛ kɛse.
Nhwɛso: Quantum computing betumi ama AI nhomasua ayɛ ntɛm sen classical limits.
Reasoning Engine (Reasoning Engine)
Reasoning Engine
Nhyehyɛe bi wɔ AI mu a ɛde adwene mu nneɛma fi nsɛm anaa data mu denam mmara anaa inference algorithms so.
Nhwɛso: AI diagnosis tool bi de reasoning engine di dwuma de hu aduruyɛ mu nsɛm a ebetumi aba denam nsɛm a ɛfa ho so.
Reinforcement Learning (RL) (Reinforcement Learning (RL))
Reinforcement Learning (RL)
Machine learning mu adwuma bi a agents sua ade denam wɔn ntam nkitahodi ne wɔn nneɛma a atwa wɔn ho ahyia no so de ma akatua no yɛ kɛse.
Nhwɛso: Robot bi a ɛresua sɛnea ɛbɛnante denam trial and error so denam RL akwan so.
Reinforcement Learning with Human Feedback (RLHF) (Reinforcement Learning with Human Feedback (RLHF))
Reinforcement Learning with Human Feedback (RLHF)
Nhomasua kwan bi a nnipa pɛ no kyerɛ AI akatua no, wɔtaa de di dwuma wɔ kasa nhwɛso ahorow a wɔyɛ no yiye mu.
Nhwɛso: Wɔkyerɛkyerɛɛ ChatGPT wɔ RLHF so de yɛ mmuae a ɛboa na ɛyɛ ahobammɔ.
Retrieval-Augmented Generation (RAG) (Retrieval-Augmented Generation (RAG))
Retrieval-Augmented Generation (RAG)
Ɔkwan bi a ɛka nsɛm a wɔhwehwɛ ne generation bom, baabi a LLM no hwehwɛ nkrataa a ɛfata de ma ne mmuae no yɛ papa.
Nhwɛso: AI boafo bi a ɛhwehwɛ na ɛkyerɛ product specs bere a ɛyɛ mmuae ma nsɛmmisa a ɛfa mfiridwuma ho.
Self-Supervised Learning (Self-Supervised Learning)
Self-Supervised Learning
Nhomasua kwan bi a nhwɛso no sua nneɛma denam ne labels a ɛyɛ no fi raw data mu, na ɛtew nnipa a wɔde kyerɛw data no so.
Nhwɛso: Wɔkyerɛkyerɛ BERT wɔ self-supervised learning so denam nsɛmfua a ɛyɛ missing wɔ text mu a wɔhyɛ ho nkɔm.
Semantic Search (Semantic Search)
Semantic Search
Hwehwɛ kwan bi a ɛte nea ɔde di dwuma no botae ne nsɛm a ɛwɔ nkyerɛase ase, ɛnyɛ keyword matching nko.
Nhwɛso: Sɛ wuhwehwɛ 'sɛnea wobɛsiesie tap a ɛtɔ' no ma guides mpo sɛ asɛmfua 'leaking tap' no nni hɔ wɔ document no mu a.
Sentiment Analysis (Sentiment Analysis)
Sentiment Analysis
Dwumadi a wɔde hu adwene, adwene, anaa suban wɔ text mu, wɔtaa kyerɛkyerɛ mu sɛ ɛyɛ papa, bɔne, anaa neutral.
Nhwɛso: Sɛ wuhwehwɛ tweets de hu ɔmanfo adwene wɔ product foforo ho.
Stochastic (Stochastic)
Stochastic
Æfa randomness anaa probabilistic suban ho, wÉ”taa de di dwuma wÉ” generative AI ne optimisation algorithms mu.
Nhwɛso: GPT-4 output no sesa ma input koro no ara esiane ne stochastic decoding process nti.
Strong AI (Strong AI)
Strong AI
Wɔsan frɛ no Artificial General Intelligence (AGI), ɛfa mfidie a ɛwɔ nnipa adwene mu tumi wɔ mmeae nyinaa ho.
Nhwɛso: AI a ɛbɛba daakye a ebetumi akyerɛw novels, ahyɛ nkurow ho nhyehyɛe, na asiesie ethical dilemmas pɛpɛɛpɛ.
Super Artificial Intelligence (SAI) (Super Artificial Intelligence (SAI))
Super Artificial Intelligence (SAI)
AI a ɛwɔ adwene mu a ɛboro nnipa adwene so wɔ nneɛma nyinaa mu — adwene, creativity, emotional intelligence, ne nea ɛkeka ho.
Nhwɛso: SAI betumi ayɛ nyansahu ne nyansapɛ foforo wɔ ne ho so.
Supervised Learning (Supervised Learning)
Supervised Learning
Machine learning kwan bi a wɔde kyerɛkyerɛ models wɔ labelled data so de sua input-output mappings.
Nhwɛso: Sɛ wode model kyerɛkyerɛ mu sɛnea ɛbɛkyerɛkyerɛ email sɛ ɛyɛ spam anaa ɛnyɛ spam denam abakɔsɛm mu nhwɛso so.
Synthetic Data (Synthetic Data)
Synthetic Data
Data a wɔayɛ no wɔ ɔkwan a ɛnteɛ so a ɛyɛ real-world data, wɔtaa de di dwuma wɔ nhomasua mu bere a real data no yɛ kakraa bi anaa ɛyɛ sensitive.
Nhwɛso: Sɛ woyɛ synthetic medical images de kyerɛkyerɛ diagnostic models a ɛmfa patient privacy nni dwuma.
Token (Token)
Token
Text unit bi a LLMs yɛ — wɔtaa yɛ asɛmfua anaa asɛmfua fã.
Nhwɛso: Asɛm 'Hello world!' no mu kyekyɛ mu tokens 3: 'Hello', 'world', ne '!'.
Tokenisation (Tokenisation)
Tokenisation
Dwumadi a wɔde text mu kyekyɛ mu tokens de yɛ adwuma wɔ model bi mu.
Nhwɛso: Wɔ NLP mu no, 'ChatGPT is great' no bɛyɛ ['Chat', 'G', 'PT', 'is', 'great'].
Transfer Learning (Transfer Learning)
Transfer Learning
Sɛ wode nimdeɛ fi adwuma bi mu di dwuma de ma nhomasua no yɛ papa wɔ adwuma foforo a ɛfa ho so, na ɛtew nhomasua bere ne data a ɛho hia no so.
Nhwɛso: Sɛ wode model a wɔakyerɛkyerɛ mu wɔ English text so yɛ sentiment analysis wɔ kasa foforo mu.
Transformer (Transformer)
Transformer
Neural network architecture bi a ɛde attention mechanisms di dwuma de yɛ sequential data, wɔtaa de di dwuma wɔ LLMs mu.
Nhwɛso: BERT, GPT, ne T5 yɛ transformer-based models.
Underfitting (Underfitting)
Underfitting
Bere a model bi yɛ mmerɛw dodo sɛ ɛbɛkora nneɛma wɔ nhomasua data mu, na ɛma ɛyɛ adwuma bɔne.
Nhwɛso: Linear model bi a ɛbɔ mmɔden sɛ ɛbɛhyɛ mfonini a emu yɛ den ho nkɔm no betumi ayɛ underfit.
Unsupervised Learning (Unsupervised Learning)
Unsupervised Learning
Nhomasua kwan bi a models hu nneɛma anaa clusters wɔ unlabelled data mu.
Nhwɛso: Sɛ wode nnipa bom denam wɔn purchasing behaviour so a labels nni hɔ.
User Intent (User Intent)
User Intent
Botae anaa atirimpɔw a ɛwɔ nea ɔde di dwuma no nsɛmmisa anaa nkitahodi akyi.
Nhwɛso: Ɔdefo bi a ɔkyerɛw 'sɛnea wobɛyɛ cake' no pɛ sɛ ohu recipe.
Validation Set (Validation Set)
Validation Set
Data fã bi a wɔde di dwuma de hwehwɛ model performance wɔ nhomasua mu na wɔyɛ hyperparameters.
Nhwɛso: Wɔde di dwuma de hu overfitting ansa na wɔayɛ test a etwa to.
Vector Database (Vector Database)
Vector Database
Database bi a wɔayɛ no sɛ ɛbɛkora na ahwehwɛ vector embeddings a wɔde di dwuma wɔ AI adwuma ahorow mu te sɛ similarity search ne RAG.
Nhwɛso: Pinecone ne Weaviate yɛ vector databases a wɔde kora text anaa image embeddings.
Vector Embedding (Vector Embedding)
Vector Embedding
Data a wɔde kyerɛw no wɔ nɔma mu a ɛkora semantic meaning ne nkitahodi wɔ vector space mu.
Nhwɛso: Nsɛmfua 'king' ne 'queen' no wɔ embeddings a ɛte sɛ nea ɛwɔ hɔ nanso ɛwɔ ɔbarima ne ɔbea nna mu nsonsonoe.
Virtual Assistant (Virtual Assistant)
Virtual Assistant
AI-powered software agent bi a ɛboa nnipa ma wɔyɛ adwuma denam kasa anaa voice commands so.
Nhwɛso: Siri, Alexa, ne Google Assistant yɛ virtual assistants a agye din.
Voice Recognition (Voice Recognition)
Voice Recognition
Mfiridwuma a ɛkyerɛkyerɛ na ɛsesa kasa kɔ text anaa adwuma mu.
Nhwɛso: Voice typing ne voice commands no fa voice recognition systems ho.
Weak AI (Weak AI)
Weak AI
AI nhyehyɛe a wɔayɛ no sɛ ɛbɛyɛ adwuma pɔtee bi a adwene nni mu.
Nhwɛso: Chess-playing AI bi a ɛntumi nte kasa ase anaa ɛntumi nni kar no yɛ weak AI nhwɛso.
Web Scraping (Web Scraping)
Web Scraping
Nsɛm a wɔde fi websites mu kɔ, wɔtaa de di dwuma de kora nhomasua data anaa wɔhwɛ nsɛm.
Nhwɛso: Sɛ wode real estate listings kora de kyerɛkyerɛ property valuation model.
Weight (Weight)
Weight
Parameter bi wɔ neural networks mu a ɛkyerɛ sɛnea node bi wɔ tumi wɔ foforo so.
Nhwɛso: Weights no sesa wɔ nhomasua mu de tew model no mfomso so.
Whisper (Whisper)
Whisper
Speech-to-text model bi a OpenAI yɛe a ebetumi akyerɛw audio wɔ kasa ahorow pii mu.
Nhwɛso: Whisper betumi akyerɛw lectures ne podcasts a ɛyɛ pɛpɛɛpɛ.
YAML (YAML)
YAML
Human-readable format bi a wɔde yɛ data serialisation, wɔtaa de di dwuma wɔ configuration files mu wɔ machine learning workflows mu.
Nhwɛso: Sɛ wode model parameters kyerɛkyerɛ mu wɔ YAML file mu de yɛ nhomasua wɔ PyTorch mu.
Zero-shot Learning (Zero-shot Learning)
Zero-shot Learning
Model bi tumi a ɛyɛ adwuma a wɔamfa nkyerɛkyerɛ mu pɛpɛɛpɛ nanso ɛde nimdeɛ a ɛyɛ foforo di dwuma.
Nhwɛso: Model bi a ɛyɛ mmuae ma mmara ho nsɛmmisa mpo sɛ wɔamfa nkyerɛkyerɛ mu pɛpɛɛpɛ wɔ mmara data so.
Zettabyte (Zettabyte)
Zettabyte
Digital data unit bi a ɛyɛ sextillion (10^21) bytes, wɔtaa de di dwuma de kyerɛ internet data kɛse.
Nhwɛso: Amansan internet traffic no boro 1 zettabyte afe biara wɔ afe 2016 mu.
