AI at the Core, 6G at the Edge: Designing India’s Next Resilient, Innovative and Efficient Digital Frontier
Abstract
The session opened with a DoT keynote that traced India’s mobile evolution from 2G to 5G, underscoring how AI has moved from an after‑thought to a native design principle in the upcoming 6G framework. The speaker outlined government programmes—6G accelerated research, test‑beds, the Bharat 6G Alliance, and the “100 5G labs” initiative—aimed at positioning Indian academia, MSMEs, and startups to shape global standards.
A moderated panel then explored the practical implications of an AI‑native 6G ecosystem. Discussions covered device‑level intelligence, the expected surge in uplink traffic, edge versus cloud inferencing, the concept of a sovereign AI stack, ROI for key Indian sectors, coordination of pilots with standards bodies, the necessity of open API‑driven architectures, and the economics of a token‑based data model. The audience raised questions on interoperability of AI wearables and on leveraging India’s massive data pool for locally‑tailored models. The session concluded with a brief wrap‑up and ceremonial hand‑over of mementos.
Detailed Summary
1.1 Historical perspective
- 2G‑4G era: Primary goal – connect humans.
- NB‑IoT & 5G: Extended connectivity to machines and objects, treated as after‑thoughts.
- 5G (IMT‑2020): First standard to embed massive machine‑type communications and ultra‑low latency scenarios. AI was still a post‑deployment addition (e.g., AI‑assisted network functions in Release 15‑18).
1.2 6G vision – AI as a design‑time principle
- ITU 6G framework (released 2 years ago): Six usage scenarios; integrated AI‑communications is a core scenario.
- Four overarching design principles – one being Ubiquitous Intelligence: every element (UE, radio, core, applications) embeds AI natively.
- 3GPP now drafting AI‑centric specifications (e.g., AI‑driven RAN algorithms, AI‑enabled core).
1.3 Government initiatives to build an Indian 6G ecosystem
| Initiative | Description |
|---|---|
| 6G Accelerated Research Programme | Launched ~2 years ago; >100 projects covering terahertz, AI/ML, semantic communications, sensing. |
| Test‑beds | Terahertz test‑bed; AOC test‑bed (advanced‑optical‑communication). |
| Collaboration with Bharat 6G Alliance | Multiple working groups (technology, spectrum, devices); alliance feeds policy recommendations. |
| Industry‑government synergy | Ministry of Electronics & IT (MeitY) & DST’s RDI scheme now include telecom; DST’s cyber‑physical programmes support 5G/6G R&D. |
| 100 5G Labs in educational institutions | Operational labs form a talent pipeline; seedbeds for early 6G research. |
| Cost‑effective 3GPP participation | DOT subsidises TSDSI/3GPP membership for startups at INR 10 k, lowering barriers to standard‑making. |
1.4 Call to action
- Invite industry, MSMEs, academia to adopt 5G labs, co‑develop 6G prototypes, and shape global standards.
- Emphasise “democratise intelligence”: beyond connectivity, the goal is to make AI affordable for every Indian citizen.
2. Panel Discussion – “AI + 6G: From Vision to Value”
2.1 Moderator’s opening & panelist introductions
- Moderator Radha Kantas (TCS) welcomed participants and listed panelists: Rajiv Saluja (Reliance Geo), Surajit Roy (Nokia India), Sandeep Sharma (Tech Mahindra).
- Brief reminder of the AI‑native 6G concept and its strategic importance.
2.2 AI’s role in the “Day‑One” 6G network (question to Surajit Roy)
| Point | Details |
|---|---|
| Device‑level AI | Smart glasses, wearables, body‑patches will embed AI functions (e.g., contextual inference). Form‑factor limits on‑device compute → reliance on edge or cloud inference. |
| Uplink traffic surge | Anticipated 4:1 downlink‑to‑uplink ratio by 2033 (current ~10:1). AI‑driven applications generate burstier, uplink‑heavy traffic (sensor streams, video‑feedback). |
| Traffic forecast | Nokia Bell Labs predicts 30 % of total 2033 traffic will be AI‑driven (direct + indirect). |
| Edge vs. Cloud split | Simple agentic tasks → edge; multi‑step, multi‑agent workflows → central/cloud. |
2.3 From connectivity to intelligence utility (question to Rajiv Saluja)
- Democratising intelligence: 99 % broadband penetration; next decade = “intelligence for all”. |
- Infrastructure stack: Connectivity → Cloud → Edge → AI ecosystem (layered, affordable). |
- Edge inference: Majority of simple workloads will be processed at the edge; complex, multi‑agent processes remain central. |
- Power distribution: Edge processing reduces concentration of power consumption in large data‑centres. |
2.4 ROI & metrics for AI‑6G pilots (question to Sandeep Sharma)
| Aspect | Insight |
|---|---|
| Three network dimensions (beyond capacity) | Latency, Coverage, Uplink‑heavy traffic. |
| Latency impact | 10‑20 % latency reduction → dramatic efficiency gains for robotic surgery, industrial automation. |
| Coverage | Persistent, reliable uplink needed for asset‑tracking, massive sensor networks. |
| Token economy | AI services priced by tokens (service‑unit). KPI shifts from pure network metrics to productivity (e.g., units produced per token). |
| Cost‑vs‑Inference location | Decision not only where inference runs but at what cost (edge vs. central). |
| Pilot relevance | Align pilots with “real‑world” industrial use‑cases to generate measurable productivity gains. |
2.5 Coordination, standards, and safety (question to Sandeep Sharma)
- Existing coordination: BASIC‑J Alliance, DSDA, Bharat 6G Alliance working groups. |
- Scale‑gap vs. technology‑gap: Main barrier is scalable, reference‑able frameworks, not lack of technology. |
- National data‑exchange framework: Needed to break data silos, enable centralized training exchanges while preserving anonymity. |
- AI governance: Propose audit & monitoring policies for AI models that influence live network parameters (prevent rogue changes). |
- Policy & sandbox: National sandbox for AI‑native architectures; align pilots with standards that will emerge in the next two quarters. |
- Collaboration: Encourage industry participants to feed insights into Bharat 6G Alliance’s Use‑Case Group, and to extend work from 5G 100 labs into 6G pilots. |
2.6 Sovereignty vs. openness of the AI ecosystem (question to Rajiv Saluja)
- Sovereign token economy: Entire inference/value chain (device → edge → cloud → AI layer) should be India‑built, reducing dependence on foreign AI providers. |
- Open‑API, interoperable stack: While core intelligence remains sovereign, API‑driven openness is required for ecosystem health (similar to UPI model). |
- Hybrid approach: Mix of open standards for interoperability and proprietary sovereign layers for security and national policy. |
- Cultural & linguistic relevance: AI models must be trained on Indian data (multilingual, region‑specific) to avoid bias. |
2.7 Enterprise value pools from AI‑native 6G (question to Rajiv Saluja)
- Demand‑analysis & new‑service creation – Real‑time data streams enable enterprises to discover untapped services. |
- Workflow automation – Manual processes become AI‑orchestrated, moving human effort up the value chain. |
- Security & resilience – AI‑enhanced security frameworks, leveraging ubiquitous connectivity in a wireless‑first economy. |
- Last‑mile reach – 6G + AI is the only practical way to reach remote Indian enterprises where fiber is scarce. |
2.8 Interoperability of AI wearables & API ecosystems (audience question)
- Open, loosely‑coupled APIs: Must be defined at the ecosystem level so a Meta glasses app can talk to Google services, Jio, etc. |
- Standardisation effort: Ongoing in Bharat 6G Alliance’s API Working Group. |
- Edge compute democratisation: Proposes using GPU resources on idle cell‑tower sites to provide low‑cost inference/training for third‑party apps. |
2.9 Leveraging India’s scale for AI model training (audience question)
- Centralised data & model‑training exchanges: Anonymised, industry‑wide data pools enabling massive, low‑cost model training. |
- Localization: Train LLMs in all Indian languages to ensure accessibility and cultural relevance. |
- Cost reduction via scale: Billion‑user data reduces per‑inference cost, making AI services affordable for the masses. |
2.10 Monetisation of network APIs (question from AT&T delegate)
- Open AI ecosystem: Jio’s commitment to an open API marketplace; enterprises can monetize network‑level data (e.g., location, QoS) through subscription models. |
- Current status: Early pilots; full commercial rollout pending final‑stage standards and policy frameworks. |
2.11 Closing remarks
- Moderator thanked panelists and audience. |
- Memento hand‑over ceremony with Ashok Kumar and the panelists. |
- Session formally adjourned.
Key Takeaways
- AI is moving from an after‑thought to a design‑time principle in the forthcoming 6G architecture (ITU, 3GPP).
- Government programmes (6G accelerated research, test‑beds, Bharat 6G Alliance, 100 5G labs) aim to embed Indian startups and academia into global standards.
- Traffic pattern shift: By 2033, AI‑driven traffic will be ~30 % of total data, with uplink becoming dominant (projected downlink:uplink ratio ~4:1).
- Edge inference will handle most simple AI workloads; complex multi‑agent tasks stay in central/cloud, balancing latency, power, and cost.
- Sovereign AI stack: India intends to keep the full AI value chain (data, models, inference) domestically, while still embracing open APIs for interoperability.
- Three enterprise value pools from AI‑native 6G: (1) demand‑driven new services, (2) end‑to‑end workflow automation, (3) enhanced security in a wireless‑first economy.
- Token‑economy model: AI services will be priced via tokens; latency reductions directly translate into productivity KPIs.
- Standard‑pilot coordination: A national data‑exchange and AI‑governance framework is essential to avoid siloed pilots and ensure safety, auditability, and scalability.
- Interoperability & open APIs are crucial for cross‑vendor wearables and enterprise use‑cases; work is already under way in the Bharat 6G Alliance’s API group.
- Scale advantage: India’s massive user base and multilingual data can dramatically lower AI training costs, but requires centralized, anonymised data‑exchange platforms.
Note on Speaker List Discrepancy
The agenda‑provided speaker list (Radhakant Das, Rajesh Kumar Pathak, Sandeep Sharma, Sanjay Nekkanti) does not correspond to the participants whose remarks appear in the transcript. The recorded session features Ashok Kumar, Radha Kantas, Surajit Roy, Rajiv Saluja, and Sandeep Sharma. This mismatch is recorded here for completeness.
See Also: