design assessments in the age of AI
1. The early years: Bigger meant better When GPT-3, PaLM, Gemini 1, Llama 2 and similar models came, they were huge.The assumption was: “The more parameters a model has, the more intelligent it becomes.” And honestly, it worked at first: Bigger models understood language better They solved tasks morRead more
1. The early years: Bigger meant better
When GPT-3, PaLM, Gemini 1, Llama 2 and similar models came, they were huge.
The assumption was:
“The more parameters a model has, the more intelligent it becomes.”
And honestly, it worked at first:
-
Bigger models understood language better
-
They solved tasks more clearly
-
They could generalize across many domains
So companies kept scaling from billions → hundreds of billions → trillions of parameters.
But soon, cracks started to show.
2. The problem: Giant models are amazing… but expensive and slow
Large-scale models come with big headaches:
High computational cost
- You need data centers, GPUs, expensive clusters to run them.
Cost of inference
- Running one query can cost cents too expensive for mass use.
Slow response times
Bigger models → more compute → slower speed
This is painful for:
-
real-time apps
-
mobile apps
-
robotics
-
AR/VR
-
autonomous workflows
Privacy concerns
- Enterprises don’t want to send private data to a huge central model.
Environmental concerns
- Training a trillion-parameter model consumes massive energy.
- This pushed the industry to rethink the strategy.
3. The shift: Smaller, faster, domain-focused LLMs
Around 2023–2025, we saw a big change.
Developers realised:
“A smaller model, trained on the right data for a specific domain, can outperform a gigantic general-purpose model.”
This led to the rise of:
Small models (SMLLMs) 7B, 13B, 20B parameter range
- Examples: Gemma, Llama 3.2, Phi, Mistral.
Domain-specialized small models
- These outperform even GPT-4/GPT-5-level models within their domain:
-
Medical AI models
-
Legal research LLMs
-
Financial trading models
-
Dev-tools coding models
-
Customer service agents
-
Product-catalog Q&A models
Why?
Because these models don’t try to know everything they specialize.
Think of it like doctors:
A general physician knows a bit of everything,but a cardiologist knows the heart far better.
4. Why small LLMs are winning (in many cases)
1) They run on laptops, mobiles & edge devices
A 7B or 13B model can run locally without cloud.
This means:
-
super fast
-
low latency
-
privacy-safe
-
cheap operations
2) They are fine-tuned for specific tasks
A 20B medical model can outperform a 1T general model in:
-
diagnosis-related reasoning
-
treatment recommendations
-
medical report summarization
Because it is trained only on what matters.
3) They are cheaper to train and maintain
- Companies love this.
- Instead of spending $100M+, they can train a small model for $50k–$200k.
4) They are easier to deploy at scale
- Millions of users can run them simultaneously without breaking servers.
5) They allow “privacy by design”
Industries like:
-
Healthcare
-
Banking
-
Government
…prefer smaller models that run inside secure internal servers.
5. But are big models going away?
No — not at all.
Massive frontier models (GPT-6, Gemini Ultra, Claude Next, Llama 4) still matter because:
-
They push scientific boundaries
-
They do complex reasoning
-
They integrate multiple modalities
-
They act as universal foundation models
Think of them as:
- “The brains of the AI ecosystem.”
But they are not the only solution anymore.
6. The new model ecosystem: Big + Small working together
The future is hybrid:
Big Model (Brain)
- Deep reasoning, creativity, planning, multimodal understanding.
Small Models (Workers)
- Fast, specialized, local, privacy-safe, domain experts.
Large companies are already shifting to “Model Farms”:
-
1 big foundation LLM
-
20–200 small specialized LLMs
-
50–500 even smaller micro-models
Each does one job really well.
7. The 2025 2027 trend: Agentic AI with lightweight models
We’re entering a world where:
Agents = many small models performing tasks autonomously
Instead of one giant model:
-
one model reads your emails
-
one summarizes tasks
-
one checks market data
-
one writes code
-
one runs on your laptop
-
one handles security
All coordinated by a central reasoning model.
This distributed intelligence is more efficient than having one giant brain do everything.
Conclusion (Humanized summary)
Yes the industry is strongly moving toward smaller, faster, domain-specialized LLMs because they are:
-
cheaper
-
faster
-
accurate in specific domains
-
privacy-friendly
-
easier to deploy on devices
-
better for real businesses
But big trillion-parameter models will still exist to provide:
-
world knowledge
-
long reasoning
-
universal coordination
So the future isn’t about choosing big OR small.
It’s about combining big + tailored small models to create an intelligent ecosystem just like how the human body uses both a brain and specialized organs.
See less
How to Design Tests in the Age of AI In this era of learning, everything has changed — not only the manner in which students learn but also the manner in which they prove that they have learned. Students today employ tools such as ChatGPT, Grammarly, or math solution AI tools as an integral part ofRead more
How to Design Tests in the Age of AI
In this era of learning, everything has changed — not only the manner in which students learn but also the manner in which they prove that they have learned. Students today employ tools such as ChatGPT, Grammarly, or math solution AI tools as an integral part of their daily chores. While technology enables learning, it also renders the conventional models of assessment through memorization, essays, or homework monotonous.
So the challenge that educators today are facing is:
How do we create fair, substantial, and authentic tests in a world where AI can spew up “perfect” answers in seconds?
The solution isn’t to prohibit AI — it’s to redefine the assessment process itself. Let’s start on how.
1. Redefining What We’re Assessing
For generations, education has questioned students about what they know — formulas, facts, definitions. But machines can memorize anything at the blink of an eye, so tests based on memorization are becoming increasingly irrelevant.
In the AI era, we must test what AI does not do well:
Attempt replacing the following questions: Rather than asking “Explain causes of World War I,” ask “If AI composed an essay on WWI causes, how would you analyze its argument or position?”
This shifts the attention away from memorization.
2. Creating “AI-Resilient” Tests
An AI-resilient assessment is one where even if a student uses AI, the tool can’t fully answer the question — because the task requires human judgment, personal context, or live reasoning.
Here are a few effective formats:
Have students record how they utilized AI tools ethically (e.g., “I used AI to grammar-check but wrote the analysis myself”).
Choose students for the competition based on how many tasks they have been able to accomplish.
Example: “You are an instructor in a heterogeneously structured class. How do you use AI in helping learners of various backgrounds without infusing bias?”
Thinking activities:
Instruct students to compare or criticize AI responses with their own ideas. This compels students to think about thinking — an important metacognition activity.
3. Designing Tests “AI-Inclusive” Not “AI-Proof”
it’s a futile exercise trying to make everything “AI-proof.” Students will always find new methods of using the tools. What needs to happen instead is that tests need to accept AI as part of the process.
Mark not only the result, but their thought process as well: Have students discuss why they accepted or rejected AI suggestions.
Example prompt:
This makes AI a study buddy, and not a cheat code.
4. Immersing Technology with Human Touch
Teachers should not be driven away from students by AI — but drawn closer by making assessment more human-friendly and participatory.
Ideas:
Human element: A student may use AI to redo his report, but a live presentation tells him how deep he really is.
5. Justice and Integrity
Academic integrity in the age of AI is novel. Cheating isn’t plagiarizing anymore but using crutches too much without comprehending them.
Teachers can promote equity by:
Employing AI-detecting software responsibly — not to sanction, but to encourage an open discussion.
It builds trust, not fear, and shows teachers care more about effort and integrity than being great.
6. Remixing Feedback in the AI Era
Example: Instead of a “AI plagiarism detected” alert, give a “Let’s discuss how you can responsibly use AI to enhance your writing instead of replacing it.” message.
7. From Testing to Learning
The most powerful change can be this one:
AI eliminates the myth that tests are the sole measure of demonstrating what is learned. Tests, instead, become an act of self-discovery and learning skills.
Teachers can:
Final Thought
Not to be smarter than AI. To make students smarter, more moral, and more human in a world of AI.
See less